- Email: [email protected]
Effect of flow maldistribution on the thermal performance of parallel microchannel cooling systems
International Journal of Heat and Mass Transfer 73 (2014) 424–428
Contents lists available at ScienceDirect
International Journal of Heat and Mass Transfer journal homepage: www.elsevier.com/locate/ijhmt
Technical Note
Effect of flow maldistribution on the thermal performance of parallel microchannel cooling systems Manoj Siva V., Arvind Pattamatta ⇑, Sarit K. Das Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai 600036, India
a r t i c l e
i n f o
Article history: Received 22 May 2013 Received in revised form 5 January 2014 Accepted 8 February 2014 Available online 14 March 2014 Keywords: Microchannel Parallel microchannel system Flow maldistribution Hot spot temperature Temperature maldistribution MEMS
a b s t r a c t This paper brings out the phenomenon of the influence of flow maldistribution on temperature distribution in parallel microchannel system that is supposed to have an adverse effect on hot spot formation in microelectronic devices. An extensive experimental study is carried out where in the parameters affecting the flow maldistribution such as channel hydraulic diameter, channel flow configurations (U, Z, I type) and chip power are varied to study their effect on the pressure drop and temperature distribution across the parallel channels designed for liquid cooling of a CPU using distilled water. It is observed that the flow distribution among the channels improves significantly with a decrease in the channel hydraulic diameter due to higher pressure drop offered by each individual channels simultaneously. This results in a considerable reduction in both the peak temperature and the average temperature of the device with decrease in channel diameter and better temperature distribution. It is observed that a higher pressure drop in d = 88 lm induces more uniform distribution compared to d = 176 lm resulting in a 3 °C improvement in the standard deviation of temperature on the chip surface and a reduction in maximum surface temperature. Higher heat fluxes induce a reduction in viscosity of the fluid resulting in higher flow maldistribution. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Due to recent trends towards miniaturization of microelectronic devices an efficient cooling system has become an inevitable necessity for effective heat dissipation and to prevent formation of local spots. According to Moore’s law, semiconductor transistor density doubles roughly every eighteen months causing higher heat load which need to be dissipated efficiently, such that for every 10 °C rise in the junction temperature, the device failure rate doubles. There are several methods of cooling available such as heat sinks, heat pipe, phase change materials and microchannel cooling systems. It is the pioneering work of Tuckerman and Pease [1] that opened up a novel cooling method using microchannel heat exchangers. Parallel microchannel cooling system stands out as a potential candidate with its higher surface area to volume ratio and higher heat dissipation rates. However, the efficiency of the parallel microchannel cooling system can be considerably reduced by non-uniform flow distribution across the microchannels called flow maldistribution resulting in the formation of localized hot spots in the device. The extent of flow maldistribution in a ⇑ Corresponding author. Tel.: +91 4422574654. E-mail address: [email protected] (A. Pattamatta). http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.02.017 0017-9310/Ó 2014 Elsevier Ltd. All rights reserved.
macrochannel (Plate Heat exchanger) is defined by the flow maldistribution parameter (m2), defined by Bassiouny and Martin [2,3] for U and Z type of flow configuration. A critical review by Kandlikar [4] stresses on the fact that characterization of flow distribution in microchannels is essential for the proper design of parallel microchannel cooling systems. Hetsroni et al. [5] studied the effect of geometry on flow and heat transfer in microchannel and observed vapor flow patterns in parallel microchannels using high speed camera. Jones et al. [6] used micro PIV technique to measure flow distribution, at different Reynolds number Re = 10 and Re = 100 on flow distribution. Seghal et al. [7] experimentally compared the effect of flow configurations such as U, P and S type on over all heat transfer and pressure drop. Tonumura et al. [8] studied the effect of length of the channel, cross sectional areas of the inlet and outlet manifolds and the shape of the manifold on flow distribution, Lu et al. [9] carried out a numerical study on parallel channel cold plate to evaluate velocity maldistribution and non-uniform distribution of temperature field caused by the fluid flow configuration. Chein et al. [10] conducted a detailed numerical study on the effect of five different inlet/outlet configurations on flow distribution through comparison of the temperature contours. Recently, Manoj Siva et al. [14,15] conducted a numerical study to understand the role of the microchannel system flow and geometric
V. Manoj Siva et al. / International Journal of Heat and Mass Transfer 73 (2014) 424–428
425
Nomenclature
l g
viscosity of the fluid, Pa s flow maldistribution factor (present case)
m Ac m2
Symbols d f l n H
hydraulic diameter of channel, lm friction factor of the channel length of the channel, mm number of channels depth of the channel, lm
Ap DPmax DPmin Wt
parameters on flow maldistribution and used optimization tools to design a system offering the best compromise in terms of overall pressure drop and flow maldistribution. The literature pertaining to flow maldistribution in microchannels as discussed above are relatively few in number and influence of different parameters on thermal performance of the microchannel has not been explicitly discussed. The causes of flow maldistribution are the ratio of channel to manifold area, number of channels, flow configuration (U, Z and I), manifold and channel wall friction, and change in flow and fluid properties. Henceforth a detailed parametric study involving channel hydraulic diameter, Power input and flow configuration are considered for the analysis of microchannel flow and temperature maldistribution. This is achieved experimentally by measuring the chip surface temperature and local variation of pressure drop across the parallel channels. The present work analyzes the effect of flow maldistribution in the context of heat transfer capability and hot spot formation in parallel microchannel cooling system.
2. Experimental description A parametric study involving the effect of the geometrical parameters such as channel hydraulic diameter/ area ratio and flow configuration for different Reynolds number have been studied. The design configurations of the present experimental study are described in Table 1. The objective of the experiment is to characterize the extent of flow maldistribution across different channel dimensions ranging from microns to the order of mm to understand flow and temperature distribution especially the size effect. Based on Kandlikar [11] channel classification, channels below 200 lm are classified as microchannels, the channel sizes chosen for the present study are 88 lm (microchannel), 176 lm (transition to micro), and 352 lm (macrochannel). The range of values for channel diameter and the number of channels are fixed to ensure that the resulting microchannel system can be accommodated on a 25 25 mm2 chip area made up of aluminium integrated with nichrome wire for heating whose resistivity is 16 O/m. The aspect ratio of the channel H/Wt is fixed at 0.1. A smaller value of the aspect ratio is chosen in order to maintain the top width of the channels larger than the hole size of 0.5 mm drilled for pressure tapping. The microchannel is fabricated through MEMS technology using well known photolithographic wet etching process as clearly
Table 1 Design configuration. d (lm)
n
Ac/Ap
Re
Chip power (W)
Configuration
88 176 352
10 10 10
0.12 0.48 1.92
10–200 10–200 10–200
5–30 5–30 5–30
U, Z, I U, Z, I U, Z, I
mass flow rate, kg/s cross sectional area of the channel, (lm)2 flow maldistribution parameter (Bassiouny and Martin model) cross sectional area of the manifold, (lm)2 maximum pressure drop minimum pressure drop top width of the channel, lm
described in (Vijayalakshmi et al. [12] and Singh et al. [13]). The details of test loop and experimental methodology are explained in Manoj Siva et al. [15]. A DC power supply 30 V, 2 A is used for heating the nichrome wire for a maximum heat flux of 5 W/cm2 approximately. The thermocouples are attached to the chip surface by proper application of thermal paste to ensure proper contact and minimize contact resistance. Calibrated K type thermocouples are used to measure chip average temperature at different locations as well as local variation of temperature and outlet temperature of the fluid which are connected to the data logger. Local variation in pressure drop and local variation in surface temperature on the chip are measured simultaneously once steady state is reached. 3. Results and discussions In this section the effect of flow maldistribution on the chip surface temperature is evaluated. Flow maldistribution is characterized based on pressure drop measured across the parallel channels using flow maldistribution parameter g. The extent of flow maldistribution is characterized based on the ratio g = (DPmax DPmin)/DPmax. A smaller value of g indicates better flow distribution and vice versa. The thermal performance of the microchannel is characterized based on the average chip surface temperature, maximum temperature and the standard deviation of the chip surface temperatures. Average temperature and maximum temperature on the chip surface quantifies the overall heat removal rate and localized heat removal rates, respectively, while standard deviation quantifies temperature non-uniformity such as thermal maldistribution which is the heat transfer equivalent to flow maldistribution parameter g. In an ideal heat removal scenario, the values of average temperature and maximum temperature are to be close to each other while the standard deviation is kept minimum. A parametric study involving the parameters such as channel hydraulic diameter, flow configuration and power input on temperature maldistribution have been evaluated and their influence on the average and peak values of temperature are reported. 3.1. Effect of channel hydraulic diameter on flow and temperature distribution In this section the influence of the channel hydraulic diameter on temperature maldistribution is discussed. As extreme diameters are four times apart, it was very difficult to maintain the mass flow rate the same for all the three cases of channel diameters; hence the experiments are facilitated in pairs, d = 88 lm and d = 176 lm compared at same mass flow rate and d = 176 lm and d = 352 lm are compared at same mass flow rate. Pressure drop across the channels, chip surface temperature are measured simultaneously to calculate flow maldistribution parameter g,
426
V. Manoj Siva et al. / International Journal of Heat and Mass Transfer 73 (2014) 424–428
direct correspondence to the average chip surface temperature and localized hot spot formation. In case of smaller hydraulic diameters of 88 lm, in which the flow maldistribution is the minimum, the resulting average temperature and standard deviation is the minimum. But in the case of larger diameter channels where flow maldistribution is high the last channels are devoid of fluid as observed in a U type flow configuration resulting in higher standard deviation and maximum temperature. It is found that flow distribution does have a direct impact on the temperature distribution hence due care needs to be given in minimizing flow distribution. 3.2. Effect of flow configuration on temperature distribution
Fig. 1a. Average, maximum temperature difference and standard deviation for U type flow configuration, m = 0.0008 kg/s, 30 W.
maximum chip surface temperature and standard deviation of chip surface temperature. The maldistribution parameter across the parallel channels is 84% for d = 176 lm and 95% for d = 352 lm for a U type flow configuration as reported in Manoj Siva et al. [15]. Thermal experiments need to be conducted at same mass flow rate to ensure that heat carried away by the different system in comparison is the same. Fig. 1a depicts the average and maximum temperature measured on the chip surface for d = 176 lm and d = 352 lm. and maximum temperature corresponds to maximum temperature on the chip surface. From Fig. 1b depicts the comparison of average and maximum temperature measured on the chip surface between the channel hydraulic diameter of d = 176 lm and d = 88 lm. The average temperature corresponds to average of the measured thermocouple data at five different locations of the chip surface. Fig. 1a it is observed that there is a significant drop in the measured maximum temperature and average by about 10 °C for d = 352 lm to d = 176 lm. The standard deviation of temperature on the chip surface is 5 °C for d = 352 lm and 3.5 °C for d = 176 lm. While from Fig. 1b it is observed that the measured maximum temperatures are about 3 °C higher for d = 176 lm, respectively, compared to d = 88 lm. Also it is observed that the standard deviation of temperature across the chip surface is 4 °C for d = 176 lm and 1 °C for d = 88 lm. This shows that the maximum temperature and standard deviation is higher d = 176 lm which also has more flow maldistribution than d = 88 lm. It can be concluded that the flow maldistribution has a
Fig. 1b. Average, maximum temperature difference and standard deviation for U type flow configuration, m = 0.00017 kg/s, 10 W.
In this section influence of flow configuration (U, Z, I type) on the temperature distribution is discussed keeping the mass flow rate and input power fixed. The values of maldistribution parameter g across the parallel channels for d = 88 lm are 84%, 54% and 28% for U, Z and I type flow configuration, respectively as reported in Manoj Siva et al. [15]. Fig. 2 compares the average, maximum and standard deviation temperature on the chip surface of d = 88 lm. It is observed that the average temperature as well as the maximum temperature on the chip surface is the highest for U type followed by Z and I type flow configuration. This can be attributed to the effect of flow maldistribution. Higher the flow maldistribution higher is the chip surface temperature and standard deviation of temperature. The approximated standard deviation calculated for d = 88 lm for U type flow configuration is 3 °C, 2 °C for Z type and 1 °C for I type. This confirms that better the flow distribution better is the temperature distribution in chip limiting the formation of localization of temperature hot spots. The probable locations of hot spot can be clearly identified as the areas of minimum DP/DPmax. That is fluid flow is minimum in the last channels of U type flow configuration, middle channels of Z type flow configuration and last channels on either sides of I type flow configuration. 3.3. Effect of chip power Figs. 3a and 3b discusses the effect of different power input (5 W, 10 W, 20 W) for d = 88 lm, n = 10 for Z type flow configuration. It is known that viscosity of the fluid (water) decreases with increase in fluid temperature. In the present case with increase in power input, chip surface temperatures increases, causing decrease in viscosity of the fluid resulting in higher mass flow rate for the same input pressure boundary condition. Increase in the total mass flow rate has resulted in an increase in the total pressure
Fig. 2. Average, maximum temperature difference and standard deviation for U, Z, I type flow configuration, d = 88 lm, m = 0.00019 kg/s, 19.08 W.
V. Manoj Siva et al. / International Journal of Heat and Mass Transfer 73 (2014) 424–428
427
across the system and decrease in individual pressure drop across occurs simultaneously such that both phenomena compensate each other for the increase in heat flux caused due to higher chip power. Fig. 3b shows the increase in flow maldistribution with respect to increase in chip power described using the flow maldistribution parameter g. The consequence of the above behaviour is the increase in standard deviation of temperature from almost 1 °C to 3 °C for the lowest and highest power inputs, respectively. Fig. 4 explains the influence of heat power input on flow maldistribution for d = 176 lm, n = 10 for three different flow configuration at the same mass flow rate. It is found that flow maldistribution increases linearly with increase in chip power. This is inline with the above discussions and holds true for flow configurations (U, Z, and I type) for both d = 88 lm and d = 176 lm. 4. Summary and conclusions Fig. 3a. Local variation of pressure drop for different input for Z type flow configuration, d = 88 lm.
Fig. 3b. Variation of total pressure drop with input power, d = 88 lm.
An experimental study has been conducted to characterize the extent of temperature non uniformity in parallel microchannel systems for liquid cooling of CPU due to flow maldistribution. A detailed parametric study involving variation in the channel hydraulic diameter, flow configuration U, Z and I type and chip power is conducted. Based upon the channel classification of Kandlikar [11], the channel diameters considered in this study are of sizes 88, 176 and 352 lm which range from micro to macrochannel. The extent of temperature maldistribution is characterized by the standard deviation of chip surface temperature. It is observed that a higher pressure drop in d = 88 lm (g = 54%) induces more uniform distribution compared to d = 176 lm (g = 84%) resulting in a 3 °C improvement in the standard deviation of temperature on the chip surface and a reduction in surface temperature. Similarly the standard deviation of temperature on the chip surface is 5 °C for d = 352 lm (g = 95%) and 3.5 °C for d = 176 lm (g = 84%). The standard deviation of temperature on the chip surface is higher for U type followed by Z and I type flow configuration because flow maldistribution is highest for U type and the least for I type. Decrease in viscosity of the fluid due to higher heat fluxes induces higher flow maldistribution. It is observed that the increase in total pressure drop across the system and decrease in individual pressure drop across occurs simultaneously such that both phenomena compensate each other for the increase in heat flux caused due to higher chip power. References
Fig. 4. Variation of flow maldistribution parameter with power input for different flow configurations, d = 176 lm.
drop across the system between inlet and outlet can increase. At the same time it is observed that the pressure drop across the individual channels has decreased with the increase in power input, such that the flow maldistribution is found to be higher at higher chip power. It is observed that the increase in total pressure drop
[1] D.B. Tuckerman, R.F. Pease, High-performance heat sinking for VLSI, IEEE Electron Device Lett. 2 (5) (1981) 126–130. [2] M.K. Bassiouny, H. Martin, Flow distribution and pressure drop in plate heat exchangers U type, Chem. Eng. Sci. 39 (4) (1983) 693–700. [3] M.K. Bassiouny, H. Martin, Flow distribution and pressure drop in plate heat exchangers Z type, Chem. Eng. Sci. 39 (4) (1983) 701–704. [4] S.G. Kandlikar, Microchannels And Minchannels – History, Terminology, Classification And Current Research Needs, First International Conference on Microchannels and Minichannels, ICMM, Rochester New York USA, 2003. [5] G. Hetsroni, A. Mosyak, Z. Segal, Non uniform temperature distribution in electronic devices cooled by flow in parallel micro-channels, IEEE Trans. Compon. Packag. Technol. 24 (2001) 16–23. [6] S.S. Seghal, K. Murugesan, S.K. Mohapatra, Experimental investigation of the effect of flow arrangements on the performance of a micro-channel heat sink, Exp. Heat Transfer 24 (2011) 215–233. [7] B.J. Jones, P. Lee, S.V. Garimella, Infrared micro-particle image velocimetry measurements and predictions of flow distribution in a microchannel heat sink, Int. J. Heat Mass Transfer 51 (2008) 1877–1887. [8] O. Tonomura, S. Tanaka, M. Noda, M. Kano, S. Hasebe, I. Hashimoto, CFD- based optimal design of manifold in plate fin microdevices, Chem. Eng. J. 101 (2004) 397–402. [9] M.C. Lu, C.C. Wang, Effect of the inlet location on the performance of parallel channel cold plate, Compo. Packag. Technol. 29 (1) (2006) 30–38. [10] R. Chein, J. Chen, Numerical study of the inlet/outlet arrangement effect on microchannel heat sink performance, Int. J. Therm. Sci. 48 (2009) 1627–1638. [11] S.G. Kandlikar, History, advances, and challenges in liquid flow and flow boiling heat transfer in microchannels, IHTC (2010).
428
V. Manoj Siva et al. / International Journal of Heat and Mass Transfer 73 (2014) 424–428
[12] K. Vijayalakshmi, K.B. Anoop, H.E. Patel, P.V. Harikrishna, T. Sundararajan, S.K. Das, Effects of compressibility and transition to turbulence on flow through microchannels, Int. J. Heat Mass Transfer 52 (2009) 2196–2204. [13] P.K. Singh, P.V. Harikrishna, T. Sundarajan, S.K. Das, Experimental and numerical investigation into the heat transfer study of nanofluids in microchannel, ASME J. Heat Transfer 133 (2011) 121701. [14] V. Manoj Siva, Arvind Pattamatta, S.K. Das, A numerical study of flow and temperature maldistribution in a parallel microchannel system for heat
removal in microelectronic devices, ASME J. Therm. Sci. Eng. Appl. 5 (2013) 041008. [15] V. Manoj Siva, Arvind Pattamatta, and S. K. Das, ‘‘Investigation on Flow Maldistribution of Parallel Microchannel Systems for Integrated Microelectronic Device Cooling,’’ Components, Packaging and Manufacturing Technology, IEEE Transactions on (Volume: PP , Issue: 99) (in press), http:// dx.doi.org/10.1109/TCPMT.2013.2284291.
Contents lists available at ScienceDirect
International Journal of Heat and Mass Transfer journal homepage: www.elsevier.com/locate/ijhmt
Technical Note
Effect of flow maldistribution on the thermal performance of parallel microchannel cooling systems Manoj Siva V., Arvind Pattamatta ⇑, Sarit K. Das Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai 600036, India
a r t i c l e
i n f o
Article history: Received 22 May 2013 Received in revised form 5 January 2014 Accepted 8 February 2014 Available online 14 March 2014 Keywords: Microchannel Parallel microchannel system Flow maldistribution Hot spot temperature Temperature maldistribution MEMS
a b s t r a c t This paper brings out the phenomenon of the influence of flow maldistribution on temperature distribution in parallel microchannel system that is supposed to have an adverse effect on hot spot formation in microelectronic devices. An extensive experimental study is carried out where in the parameters affecting the flow maldistribution such as channel hydraulic diameter, channel flow configurations (U, Z, I type) and chip power are varied to study their effect on the pressure drop and temperature distribution across the parallel channels designed for liquid cooling of a CPU using distilled water. It is observed that the flow distribution among the channels improves significantly with a decrease in the channel hydraulic diameter due to higher pressure drop offered by each individual channels simultaneously. This results in a considerable reduction in both the peak temperature and the average temperature of the device with decrease in channel diameter and better temperature distribution. It is observed that a higher pressure drop in d = 88 lm induces more uniform distribution compared to d = 176 lm resulting in a 3 °C improvement in the standard deviation of temperature on the chip surface and a reduction in maximum surface temperature. Higher heat fluxes induce a reduction in viscosity of the fluid resulting in higher flow maldistribution. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Due to recent trends towards miniaturization of microelectronic devices an efficient cooling system has become an inevitable necessity for effective heat dissipation and to prevent formation of local spots. According to Moore’s law, semiconductor transistor density doubles roughly every eighteen months causing higher heat load which need to be dissipated efficiently, such that for every 10 °C rise in the junction temperature, the device failure rate doubles. There are several methods of cooling available such as heat sinks, heat pipe, phase change materials and microchannel cooling systems. It is the pioneering work of Tuckerman and Pease [1] that opened up a novel cooling method using microchannel heat exchangers. Parallel microchannel cooling system stands out as a potential candidate with its higher surface area to volume ratio and higher heat dissipation rates. However, the efficiency of the parallel microchannel cooling system can be considerably reduced by non-uniform flow distribution across the microchannels called flow maldistribution resulting in the formation of localized hot spots in the device. The extent of flow maldistribution in a ⇑ Corresponding author. Tel.: +91 4422574654. E-mail address: [email protected] (A. Pattamatta). http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.02.017 0017-9310/Ó 2014 Elsevier Ltd. All rights reserved.
macrochannel (Plate Heat exchanger) is defined by the flow maldistribution parameter (m2), defined by Bassiouny and Martin [2,3] for U and Z type of flow configuration. A critical review by Kandlikar [4] stresses on the fact that characterization of flow distribution in microchannels is essential for the proper design of parallel microchannel cooling systems. Hetsroni et al. [5] studied the effect of geometry on flow and heat transfer in microchannel and observed vapor flow patterns in parallel microchannels using high speed camera. Jones et al. [6] used micro PIV technique to measure flow distribution, at different Reynolds number Re = 10 and Re = 100 on flow distribution. Seghal et al. [7] experimentally compared the effect of flow configurations such as U, P and S type on over all heat transfer and pressure drop. Tonumura et al. [8] studied the effect of length of the channel, cross sectional areas of the inlet and outlet manifolds and the shape of the manifold on flow distribution, Lu et al. [9] carried out a numerical study on parallel channel cold plate to evaluate velocity maldistribution and non-uniform distribution of temperature field caused by the fluid flow configuration. Chein et al. [10] conducted a detailed numerical study on the effect of five different inlet/outlet configurations on flow distribution through comparison of the temperature contours. Recently, Manoj Siva et al. [14,15] conducted a numerical study to understand the role of the microchannel system flow and geometric
V. Manoj Siva et al. / International Journal of Heat and Mass Transfer 73 (2014) 424–428
425
Nomenclature
l g
viscosity of the fluid, Pa s flow maldistribution factor (present case)
m Ac m2
Symbols d f l n H
hydraulic diameter of channel, lm friction factor of the channel length of the channel, mm number of channels depth of the channel, lm
Ap DPmax DPmin Wt
parameters on flow maldistribution and used optimization tools to design a system offering the best compromise in terms of overall pressure drop and flow maldistribution. The literature pertaining to flow maldistribution in microchannels as discussed above are relatively few in number and influence of different parameters on thermal performance of the microchannel has not been explicitly discussed. The causes of flow maldistribution are the ratio of channel to manifold area, number of channels, flow configuration (U, Z and I), manifold and channel wall friction, and change in flow and fluid properties. Henceforth a detailed parametric study involving channel hydraulic diameter, Power input and flow configuration are considered for the analysis of microchannel flow and temperature maldistribution. This is achieved experimentally by measuring the chip surface temperature and local variation of pressure drop across the parallel channels. The present work analyzes the effect of flow maldistribution in the context of heat transfer capability and hot spot formation in parallel microchannel cooling system.
2. Experimental description A parametric study involving the effect of the geometrical parameters such as channel hydraulic diameter/ area ratio and flow configuration for different Reynolds number have been studied. The design configurations of the present experimental study are described in Table 1. The objective of the experiment is to characterize the extent of flow maldistribution across different channel dimensions ranging from microns to the order of mm to understand flow and temperature distribution especially the size effect. Based on Kandlikar [11] channel classification, channels below 200 lm are classified as microchannels, the channel sizes chosen for the present study are 88 lm (microchannel), 176 lm (transition to micro), and 352 lm (macrochannel). The range of values for channel diameter and the number of channels are fixed to ensure that the resulting microchannel system can be accommodated on a 25 25 mm2 chip area made up of aluminium integrated with nichrome wire for heating whose resistivity is 16 O/m. The aspect ratio of the channel H/Wt is fixed at 0.1. A smaller value of the aspect ratio is chosen in order to maintain the top width of the channels larger than the hole size of 0.5 mm drilled for pressure tapping. The microchannel is fabricated through MEMS technology using well known photolithographic wet etching process as clearly
Table 1 Design configuration. d (lm)
n
Ac/Ap
Re
Chip power (W)
Configuration
88 176 352
10 10 10
0.12 0.48 1.92
10–200 10–200 10–200
5–30 5–30 5–30
U, Z, I U, Z, I U, Z, I
mass flow rate, kg/s cross sectional area of the channel, (lm)2 flow maldistribution parameter (Bassiouny and Martin model) cross sectional area of the manifold, (lm)2 maximum pressure drop minimum pressure drop top width of the channel, lm
described in (Vijayalakshmi et al. [12] and Singh et al. [13]). The details of test loop and experimental methodology are explained in Manoj Siva et al. [15]. A DC power supply 30 V, 2 A is used for heating the nichrome wire for a maximum heat flux of 5 W/cm2 approximately. The thermocouples are attached to the chip surface by proper application of thermal paste to ensure proper contact and minimize contact resistance. Calibrated K type thermocouples are used to measure chip average temperature at different locations as well as local variation of temperature and outlet temperature of the fluid which are connected to the data logger. Local variation in pressure drop and local variation in surface temperature on the chip are measured simultaneously once steady state is reached. 3. Results and discussions In this section the effect of flow maldistribution on the chip surface temperature is evaluated. Flow maldistribution is characterized based on pressure drop measured across the parallel channels using flow maldistribution parameter g. The extent of flow maldistribution is characterized based on the ratio g = (DPmax DPmin)/DPmax. A smaller value of g indicates better flow distribution and vice versa. The thermal performance of the microchannel is characterized based on the average chip surface temperature, maximum temperature and the standard deviation of the chip surface temperatures. Average temperature and maximum temperature on the chip surface quantifies the overall heat removal rate and localized heat removal rates, respectively, while standard deviation quantifies temperature non-uniformity such as thermal maldistribution which is the heat transfer equivalent to flow maldistribution parameter g. In an ideal heat removal scenario, the values of average temperature and maximum temperature are to be close to each other while the standard deviation is kept minimum. A parametric study involving the parameters such as channel hydraulic diameter, flow configuration and power input on temperature maldistribution have been evaluated and their influence on the average and peak values of temperature are reported. 3.1. Effect of channel hydraulic diameter on flow and temperature distribution In this section the influence of the channel hydraulic diameter on temperature maldistribution is discussed. As extreme diameters are four times apart, it was very difficult to maintain the mass flow rate the same for all the three cases of channel diameters; hence the experiments are facilitated in pairs, d = 88 lm and d = 176 lm compared at same mass flow rate and d = 176 lm and d = 352 lm are compared at same mass flow rate. Pressure drop across the channels, chip surface temperature are measured simultaneously to calculate flow maldistribution parameter g,
426
V. Manoj Siva et al. / International Journal of Heat and Mass Transfer 73 (2014) 424–428
direct correspondence to the average chip surface temperature and localized hot spot formation. In case of smaller hydraulic diameters of 88 lm, in which the flow maldistribution is the minimum, the resulting average temperature and standard deviation is the minimum. But in the case of larger diameter channels where flow maldistribution is high the last channels are devoid of fluid as observed in a U type flow configuration resulting in higher standard deviation and maximum temperature. It is found that flow distribution does have a direct impact on the temperature distribution hence due care needs to be given in minimizing flow distribution. 3.2. Effect of flow configuration on temperature distribution
Fig. 1a. Average, maximum temperature difference and standard deviation for U type flow configuration, m = 0.0008 kg/s, 30 W.
maximum chip surface temperature and standard deviation of chip surface temperature. The maldistribution parameter across the parallel channels is 84% for d = 176 lm and 95% for d = 352 lm for a U type flow configuration as reported in Manoj Siva et al. [15]. Thermal experiments need to be conducted at same mass flow rate to ensure that heat carried away by the different system in comparison is the same. Fig. 1a depicts the average and maximum temperature measured on the chip surface for d = 176 lm and d = 352 lm. and maximum temperature corresponds to maximum temperature on the chip surface. From Fig. 1b depicts the comparison of average and maximum temperature measured on the chip surface between the channel hydraulic diameter of d = 176 lm and d = 88 lm. The average temperature corresponds to average of the measured thermocouple data at five different locations of the chip surface. Fig. 1a it is observed that there is a significant drop in the measured maximum temperature and average by about 10 °C for d = 352 lm to d = 176 lm. The standard deviation of temperature on the chip surface is 5 °C for d = 352 lm and 3.5 °C for d = 176 lm. While from Fig. 1b it is observed that the measured maximum temperatures are about 3 °C higher for d = 176 lm, respectively, compared to d = 88 lm. Also it is observed that the standard deviation of temperature across the chip surface is 4 °C for d = 176 lm and 1 °C for d = 88 lm. This shows that the maximum temperature and standard deviation is higher d = 176 lm which also has more flow maldistribution than d = 88 lm. It can be concluded that the flow maldistribution has a
Fig. 1b. Average, maximum temperature difference and standard deviation for U type flow configuration, m = 0.00017 kg/s, 10 W.
In this section influence of flow configuration (U, Z, I type) on the temperature distribution is discussed keeping the mass flow rate and input power fixed. The values of maldistribution parameter g across the parallel channels for d = 88 lm are 84%, 54% and 28% for U, Z and I type flow configuration, respectively as reported in Manoj Siva et al. [15]. Fig. 2 compares the average, maximum and standard deviation temperature on the chip surface of d = 88 lm. It is observed that the average temperature as well as the maximum temperature on the chip surface is the highest for U type followed by Z and I type flow configuration. This can be attributed to the effect of flow maldistribution. Higher the flow maldistribution higher is the chip surface temperature and standard deviation of temperature. The approximated standard deviation calculated for d = 88 lm for U type flow configuration is 3 °C, 2 °C for Z type and 1 °C for I type. This confirms that better the flow distribution better is the temperature distribution in chip limiting the formation of localization of temperature hot spots. The probable locations of hot spot can be clearly identified as the areas of minimum DP/DPmax. That is fluid flow is minimum in the last channels of U type flow configuration, middle channels of Z type flow configuration and last channels on either sides of I type flow configuration. 3.3. Effect of chip power Figs. 3a and 3b discusses the effect of different power input (5 W, 10 W, 20 W) for d = 88 lm, n = 10 for Z type flow configuration. It is known that viscosity of the fluid (water) decreases with increase in fluid temperature. In the present case with increase in power input, chip surface temperatures increases, causing decrease in viscosity of the fluid resulting in higher mass flow rate for the same input pressure boundary condition. Increase in the total mass flow rate has resulted in an increase in the total pressure
Fig. 2. Average, maximum temperature difference and standard deviation for U, Z, I type flow configuration, d = 88 lm, m = 0.00019 kg/s, 19.08 W.
V. Manoj Siva et al. / International Journal of Heat and Mass Transfer 73 (2014) 424–428
427
across the system and decrease in individual pressure drop across occurs simultaneously such that both phenomena compensate each other for the increase in heat flux caused due to higher chip power. Fig. 3b shows the increase in flow maldistribution with respect to increase in chip power described using the flow maldistribution parameter g. The consequence of the above behaviour is the increase in standard deviation of temperature from almost 1 °C to 3 °C for the lowest and highest power inputs, respectively. Fig. 4 explains the influence of heat power input on flow maldistribution for d = 176 lm, n = 10 for three different flow configuration at the same mass flow rate. It is found that flow maldistribution increases linearly with increase in chip power. This is inline with the above discussions and holds true for flow configurations (U, Z, and I type) for both d = 88 lm and d = 176 lm. 4. Summary and conclusions Fig. 3a. Local variation of pressure drop for different input for Z type flow configuration, d = 88 lm.
Fig. 3b. Variation of total pressure drop with input power, d = 88 lm.
An experimental study has been conducted to characterize the extent of temperature non uniformity in parallel microchannel systems for liquid cooling of CPU due to flow maldistribution. A detailed parametric study involving variation in the channel hydraulic diameter, flow configuration U, Z and I type and chip power is conducted. Based upon the channel classification of Kandlikar [11], the channel diameters considered in this study are of sizes 88, 176 and 352 lm which range from micro to macrochannel. The extent of temperature maldistribution is characterized by the standard deviation of chip surface temperature. It is observed that a higher pressure drop in d = 88 lm (g = 54%) induces more uniform distribution compared to d = 176 lm (g = 84%) resulting in a 3 °C improvement in the standard deviation of temperature on the chip surface and a reduction in surface temperature. Similarly the standard deviation of temperature on the chip surface is 5 °C for d = 352 lm (g = 95%) and 3.5 °C for d = 176 lm (g = 84%). The standard deviation of temperature on the chip surface is higher for U type followed by Z and I type flow configuration because flow maldistribution is highest for U type and the least for I type. Decrease in viscosity of the fluid due to higher heat fluxes induces higher flow maldistribution. It is observed that the increase in total pressure drop across the system and decrease in individual pressure drop across occurs simultaneously such that both phenomena compensate each other for the increase in heat flux caused due to higher chip power. References
Fig. 4. Variation of flow maldistribution parameter with power input for different flow configurations, d = 176 lm.
drop across the system between inlet and outlet can increase. At the same time it is observed that the pressure drop across the individual channels has decreased with the increase in power input, such that the flow maldistribution is found to be higher at higher chip power. It is observed that the increase in total pressure drop
[1] D.B. Tuckerman, R.F. Pease, High-performance heat sinking for VLSI, IEEE Electron Device Lett. 2 (5) (1981) 126–130. [2] M.K. Bassiouny, H. Martin, Flow distribution and pressure drop in plate heat exchangers U type, Chem. Eng. Sci. 39 (4) (1983) 693–700. [3] M.K. Bassiouny, H. Martin, Flow distribution and pressure drop in plate heat exchangers Z type, Chem. Eng. Sci. 39 (4) (1983) 701–704. [4] S.G. Kandlikar, Microchannels And Minchannels – History, Terminology, Classification And Current Research Needs, First International Conference on Microchannels and Minichannels, ICMM, Rochester New York USA, 2003. [5] G. Hetsroni, A. Mosyak, Z. Segal, Non uniform temperature distribution in electronic devices cooled by flow in parallel micro-channels, IEEE Trans. Compon. Packag. Technol. 24 (2001) 16–23. [6] S.S. Seghal, K. Murugesan, S.K. Mohapatra, Experimental investigation of the effect of flow arrangements on the performance of a micro-channel heat sink, Exp. Heat Transfer 24 (2011) 215–233. [7] B.J. Jones, P. Lee, S.V. Garimella, Infrared micro-particle image velocimetry measurements and predictions of flow distribution in a microchannel heat sink, Int. J. Heat Mass Transfer 51 (2008) 1877–1887. [8] O. Tonomura, S. Tanaka, M. Noda, M. Kano, S. Hasebe, I. Hashimoto, CFD- based optimal design of manifold in plate fin microdevices, Chem. Eng. J. 101 (2004) 397–402. [9] M.C. Lu, C.C. Wang, Effect of the inlet location on the performance of parallel channel cold plate, Compo. Packag. Technol. 29 (1) (2006) 30–38. [10] R. Chein, J. Chen, Numerical study of the inlet/outlet arrangement effect on microchannel heat sink performance, Int. J. Therm. Sci. 48 (2009) 1627–1638. [11] S.G. Kandlikar, History, advances, and challenges in liquid flow and flow boiling heat transfer in microchannels, IHTC (2010).
428
V. Manoj Siva et al. / International Journal of Heat and Mass Transfer 73 (2014) 424–428
[12] K. Vijayalakshmi, K.B. Anoop, H.E. Patel, P.V. Harikrishna, T. Sundararajan, S.K. Das, Effects of compressibility and transition to turbulence on flow through microchannels, Int. J. Heat Mass Transfer 52 (2009) 2196–2204. [13] P.K. Singh, P.V. Harikrishna, T. Sundarajan, S.K. Das, Experimental and numerical investigation into the heat transfer study of nanofluids in microchannel, ASME J. Heat Transfer 133 (2011) 121701. [14] V. Manoj Siva, Arvind Pattamatta, S.K. Das, A numerical study of flow and temperature maldistribution in a parallel microchannel system for heat
removal in microelectronic devices, ASME J. Therm. Sci. Eng. Appl. 5 (2013) 041008. [15] V. Manoj Siva, Arvind Pattamatta, and S. K. Das, ‘‘Investigation on Flow Maldistribution of Parallel Microchannel Systems for Integrated Microelectronic Device Cooling,’’ Components, Packaging and Manufacturing Technology, IEEE Transactions on (Volume: PP , Issue: 99) (in press), http:// dx.doi.org/10.1109/TCPMT.2013.2284291.