INTERNATIONAL JOURNAL OF ENERGY RESEARCH Int. J. Energy Res.

2007; 31:717727 Published online 11 December 2006 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/er.1287

Design, fabrication and thermal characterization of a magnetocaloric microcooler

S. Kim, S. Ghirlanda, C. Adams, B. Bethala, S. N. Sambandam and S. Bhansali*,y
BioMEMS and Microsystems Laboratory, Department of Electrical Engineering, University of South Florida, 4202 E. Fowler Ave., ENB118, Tampa, FL 33620, U.S.A.

SUMMARY Magnetocaloric cooling is an alternative, high-eciency cooling technology. In this paper, we present the design and fabrication of a micromachined magnetocaloric cooler and demonstrate its ability to work in a small magnetic eld 51:2 T with a cooling test. The cooler was built by fabricating Si microuidic channels, and it was integrated with a Gd5 Si2 Ge2 magnetocaloric refrigeration element. The magnetic properties of the Gd5 Si2 Ge2 material were characterized to calculate the magnetic entropy change at dierent ambient temperatures. Three dierent methods to integrate the channel layer and the magnetocaloric element were evaluated to test sealing and cooling performance. The cooling tests were performed by providing a magnetic eld using an electromagnet. A test jig was constructed between the poles of an electromagnet to maintain a steady temperature during the test. Cooling tests were performed on the magnetocaloric element at ambient temperatures ranging from 258 to 280 K using a magnetic eld of 1.2 T. Experimental results showed a maximum temperature change of 7 K on the magnetocaloric element alone at an ambient temperature of 258 K. Cooling tests of the fully integrated coolers were also performed. A solution of anti-freeze uid (propylene glycol) and water was used as the coolant. The temperature of the working uid decreased by 4.6 and 9 K for the glass and Si intermediate layers, respectively, conrming that the thermal conductivity of the materials is also an important factor in cooler performance. Copyright # 2006 John Wiley & Sons, Ltd.
KEY WORDS:

magnetocaloric; microcooling; Si microchannel; Gd5 Si2 Ge2 ; magnetocaloric entropy; microuidic

*Correspondence to: S. Bhansali, Department of Electrical Engineering, University of South Florida, 4202 E. Fowler Ave., ENB118, Tampa, FL 33620, U.S.A. y E-mail: bhansali@eng.usf.edu Contract/grant sponsor: NASA Received 1 January 2006 Accepted 18 October 2006

Copyright # 2006 John Wiley & Sons, Ltd.

718

S. KIM ET AL.

1. INTRODUCTION Magnetocaloric refrigeration (Pecharsky and Gschneidner, 1999a; Bohigas et al., 2000) is a promising alternative technique for low-temperature cooling, which is highly ecient > 60%; compact in size, and environmentally friendly (does not use coolants or emit greenhouse gases). Unlike conventional refrigerators that use the expansion and compression of a refrigerant, the magnetocaloric refrigerator uses cooling elements made of magnetocaloric materials. Magnetocaloric materials, when subjected to a magnetic eld at their phase transition temperature, undergo a magnetic transition which results in the entropy change of the magnetocaloric materials (Pecharsky and Gschneidner, 1999a). This phenomenon is known as the magnetocaloric eect (MCE). This entropy change results in a temperature change that can be used to heat/cool the coolant. Amongst the dierent materials that exhibit MCE, the Gd5 Six Ge1x 4 family has shown the highest MCE, especially with x40:5 (Pecharsky and Gschneidner, 1999a). In this research, an alloy with a target composition of Gd5 Si2 Ge2 x 0:5 was used. Magnetocaloric refrigerators have been developed as a potential replacement of conventional gas compression/expansion-based refrigeration. Pecharsky and Gschneidner (1999a) presented a reciprocating magnetic refrigerator using Gd refrigerant and water as the heat transfer uid. The magnetic eld in this refrigerator was created using a superconducting magnet. Bohigas et al. (2000) proposed a room temperature magnetic refrigerator based on adiabatic magnetic refrigeration. Their experiment employed permanent magnets to generate the magnetic eld, and a 1.6 K temperature dierence between the hot and cold region was obtained at a 0.3 T magnetic eld. Shirron et al. (2002) presented an adiabatic demagnetization refrigerator that produced continuous cooling at sub-Kelvin (60 mK) temperatures. An analysis of the thermodynamic characteristics of a magnetic refrigeration cycle, consisting of a heat exchanger and magnetic beds, has been presented by Rahman and Rosario (2004). This paper explores the microfabrication of a magnetocaloric refrigerator. Miniaturized coolers have been a subject of intense research and development in the recent past (Pekola et al., 2000; Zeng et al., 2001; Kim et al., 2005). Developing coolers at a micro-scale that can operate at low temperatures continues to be a challenge. This paper presents the feasibility of magnetocaloric cooling using microuidic structures. The design, fabrication, and test results of the microcooler and its thermal characterization will be presented.

2. DESIGN AND FABRICATION The magnetocaloric microcooler shown in Figure 1 consists of Si uidic microchannels, a magnetocaloric element and an electromagnet. A Gd5 Si2 Ge2 disc is used as the magnetocaloric element. Once the cooler is integrated, a working uid is pumped from a reservoir into the cooler by a syringe pump. When the electromagnet turns on, the magnetocaloric element undergoes a magnetically driven crystallographic transformation, thereby changing its heat capacity. This results in the heating of the element. When the eld is turned o, the element cools. This heating and cooling of the element coupled with the change of the ow direction (Figure 2) draw heat from the uid. A miniature magnetocaloric refrigerator can be constructed using two such microcoolers (magnetocaloric beds) with two heat exchangers, a pump and several valves, as shown in Figure 2 (Rahman and Rosario, 2004). In the proposed
Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

MAGNETOCALORIC MICROCOOLER

719

Figure 1. Schematic diagram of the microcooler.

Figure 2. Conguration of a magnetic refrigerator (adapted from Rahman and Rosario (2004) by courtesy of authors).

system, the microcoolers are analogous to the expansion stage and compression stage of a gas refrigeration system. The microcooler was built by attaching the magnetocaloric element to a micromachined Si substrate. Eight uidic microchannels of trapezoidal shape, each 300 mm wide, 150 mm deep and 25 mm long were etched in a 1 in 1 in area on the 2-in silicon 1 0 0 wafer. The microchannels were tied together at the inlet and outlet to ensure uniform ow. The channels were placed at a pitch of 2.7 mm. The microchannel spacing was designed to ensure that an adequate mass of magnetocaloric element, Gd5 Si2 Ge2 ; was dedicated to each cooling channel. Figure 3 shows the fabrication process of microchannels. The process starts with a 500 mm thick
Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

720

S. KIM ET AL.

Figure 3. Fabrication process for microchannels in a 1 0 0 Si wafer.

double-side polished Si wafer. The wafer was then RCA cleaned, followed by the growth of a 300 nm thick thermal SiO2 layer (Figure 3(a)). Next, a negative photoresist was spun on the front side of the wafer, and the inlet and outlet ports were patterned with lithographic techniques. The back side SiO2 was protected by spinning the photoresist and hard baking it without patterning. This was then followed by a buered oxide etch (BOE) step to remove SiO2 from the exposed areas. The photoresist was then stripped from the wafer using acetone (Figure 3(b)). The wafer was subsequently wet etched in 25% tetramethylammonium hydroxide (TMAH). The TMAH was heated to 858C and a timed etch of 350 mm was performed, as shown in Figure 3(c) at the AA0 cross-section. Next, the wafer was rinsed and prepared to pattern and etch the microchannels. The front side of the wafer was again subjected to lithography using the microchannel mask. The backside SiO2 of the wafer was again protected by applying a blanket photoresist layer. The resist was developed and the oxide was etched to transfer the pattern of the microchannel into the front oxide. The resist was then stripped (Figure 3(d), BB0 crosssection). A timed wet etch was then performed for 150 mm to etch the microchannels (Figure 3e1 ; BB0 cross-section) and the inlet and outlet ports (Figure 3(e2 ; CC0 cross-section). Then, the oxide was stripped. This results in through holes at the inlet and outlet. The wet etching of 1 0 0 Si wafer results in a 54:748 side wall angle, thereby creating microchannels which have trapezoidal cross-sections. The mask had no convex corner compensation structure. This resulted in a rounding of the corners at the intersection of channels. The rounding of the corners provides a more uniform ow prole with a reduced pressure drop. As discussed earlier,
Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

MAGNETOCALORIC MICROCOOLER

721

Gd5 Si2 Ge2 was chosen as the magnetocaloric material. The Gd5 Si2 Ge2 alloy was synthesized at the AMES Laboratory, Iowa State University. The alloy was synthesized by preparing stoichiometric mixtures of Gd (99.8% purity), Si and Ge (both 99.99% purity). The fabricated Si chip with eight microchannels is shown in Figure 4. Microconnectors were attached to the inlet and outlet ports on the backside of the wafer. The front side of the wafer was then attached to the 2 in diameter, 5 mm thick Gd5 Si2 Ge2 disc using a commercial adhesive crystal bondTM 509. The adhesive was dissolved in acetone and sprayed on the Gd5 Si2 Ge2 surface. The acetone was then allowed to evaporate, resulting in a 35 mm thin island of adhesive, disbursed over the wafer (Oh et al., 2002). The wafer was then placed on the adhesive-coated disc upside down and heated on the hot plate to a temperature of 393 K, under a light pressure. Then the heater was turned o. The heating results in melting and reow of the crystal bond. As the adhesive cools, the disc bonds with the microchannels. The integrated cooler with microuidic interconnections is shown in Figure 5. In addition to crystal bonding, two dierent integration methods were also used with an objective of making a leak-proof seal on the cooler. Leak proong was a challenge due to sub-zero operating temperatures. In the rst instance, a glass wafer was bonded to the channel wafer through anodic bonding. The Si wafer with microchannels and the glass wafer were aligned and heated on a hot plate to 5008C: The wafers were biased once hot, and a high voltage of 1.5 kV was supplied to the wafers for 20 min. The stack was then cooled to room temperature. In the second approach, the Si wafer with microchannels was bonded with another Si wafer through fusion bonding. The channel wafer and an additional blank Si wafer were cleaned using an RCA process, brought into contact with each other and subjected to pressure for prebonding. The pre-bonded wafers were annealed at 10008C in the furnace in an N2 environment. The bonded wafers were then assembled with the Gd5 Si2 Ge2 block without any intermediate

Figure 4. Eight microchannels fabricated on a Si wafer.

Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

722

S. KIM ET AL.

bonding material and the edges were glued with an epoxy. After integrating the cooler, sealing tests were performed by supplying the working liquid.

3. EXPERIMENTAL SET-UP The overall schematic diagram of the cooling test set-up is shown in Figure 6. Cooling tests were performed inside a 1-in thick polystyrene Styrofoam chamber, built around the electromagnet poles. Two aluminium plates sealed next to each side of the poles inside the chamber created

Figure 5. Photograph of the integrated cooler.

Figure 6. Schematic diagram of the cooling test set-up.

Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

MAGNETOCALORIC MICROCOOLER

723

reservoirs for the liquid N2 : The temperature inside the chamber was controlled through the amount of liquid N2 in the chamber. Two aluminium plates, three-eighths inches thick, were placed in these baths of liquid N2 and bent to cover the top of the poles, thereby creating a uniform temperature gradient in the centre of the chamber where the sample rested. To ensure independently calibrated measurements, the temperature change was measured using two external thermocouples at 0:18C resolution, with data logged every second. The 1.2 T magnetic eld was generated through an electromagnet.

4. RESULTS AND DISCUSSION 4.1. Magnetocaloric properties The synthesized powders of Gd5 Si2 Ge2 were characterized for magnetic and magnetocaloric properties. The MCE eect was studied by the measurement of magnetization and calculation of the magnetic entropy change. The magnetization isotherms of Figure 7(a) were measured at the AMES Laboratory, Iowa State University for indirect calculations of MCEs. The 21 curves of magnetization and magnetic elds shown in Figure 7(a) were obtained at constant temperatures from 240.0 to 340.0 K at an interval of 5 K. The magnetization data at each temperature were obtained by changing the magnetic eld from 0 to 5 T with a step of 0.2 T. The magnetocaloric entropy change can be indirectly calculated from the magnetization data, as shown in Figure 7(a), which is related to values of the magnetic eld H ; temperature T ; and magnetization M : The following simplied equation can be used to calculate the magnetic entropy change DSM Tav DH for an average temperature of Tav Tu Tl =2 between two isotherms and a magnetic eld dierence, DH HF HI ; between the nal and initial magnetic eld (Pecharsky and Gschneidner, 1999b). ! n1 X dH dM 1 2 DSm Tav DH dM k dM n 1 2dT k2

Figure 7. Characteristics of the magnetocaloric Gd5 Si2 Ge2 material: (a) magnetization isotherms in Gd5 Si2 Ge2 ; and (b) magnetocaloric entropy change of the Gd5 Si2 Ge2 material calculated from magnetization data shown in (a).
Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

724

S. KIM ET AL.

In Equation (1), dH DH =n 1 is a constant incremental step of the magnetic eld and n is the number of measurement points within HI and HF in a constant dH : Also, dT Tu Tl is the temperature dierence between two isotherms and dMk M Tu k M Tl k is the dierence in the magnetization at Tu and Tl for each magnetic-eld step from 1 to n: Figure 7(b) shows the calculated magnetic entropy change with temperature, indicating a maximum magnetic entropy change. This gure shows two distinct peaks, which are indicative of two phases that have their distinct transition temperatures. The Gd5 Si2 Ge2 material used is not of high quality. It is mainly used to show the feasibility of the proposed magnetocaloric cooler. The MCE variation in a good-quality Gd5 Si2 Ge2 material is quite dierent from the plot shown in Figure 7(b). 4.2. Cooling characteristics of only magnetocaloric block Initially, the cooling test was performed on a Gd5 Si2 Ge2 block only. One external thermocouple was attached on the surface of the Gd5 Si2 Ge2 block and the other external thermocouple was positioned for measurement of the chamber temperature. The Gd5 Si2 Ge2 block insulated with one-half inch thick polystyrene Styrofoam was vertically positioned at the centre between the two magnetic poles. After the temperatures of both the Gd5 Si2 Ge2 block and the chamber were equalized, the test was started. The starting temperature of the block was controlled to the desired value, between 255 and 280 K using liquid nitrogen. The magnetic eld was then applied to the Gd5 Si2 Ge2 block for 90 s. Figure 8 shows the temperature changes for various initial temperatures, when the magnet is switched on. At this moment, the magnetic spins are aligned with the application of the magnetic eld, which reduces the entropy of the Gd5 Si2 Ge2 material. As seen in Figure 8, the maximum temperature change was observed at 258 K. It was also observed that as the initial temperature

Figure 8. Change in temperature at various initial ambient temperatures of the Gd5 Si2 Ge2 magnetocaloric element.
Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

MAGNETOCALORIC MICROCOOLER

725

increased, the temperature change decreased. These results can be correlated with the entropy change (Figure 7(b)). The entropy change peak, shown in Figure 7(b), exists at 262 K. The maximum temperature change of the block is observed around that peak in Figure 7. Though an entropy change peak exists at 280 K, no appreciable temperature change was observed. This may be attributed to system noise. 4.3. Cooling results of the integrated coolers 4.3.1. Working condition and set-up. Cooling tests of the coolers were performed on the test setup, as shown in Figure 6. In order to pump the uid into the channels, a syringe pump was used. The syringe plunger was moved within the syringe barrel by a stepper motor. The syringe drive has a 30 mm travel length. In order to prevent excessive pressure, which ultimately could cause leaks, a lower ow rate of 33:3 ml s1 was chosen. The working uid consisted of both anti-freeze uid and water in a concentration of 50% inhibited propylene glycol and 50% water. Once the system temperature has stabilized, the solution of water and anti-freeze would ow through the system for about 10 s, in order to eliminate all residual liquid left into the tubes and microchannels, whose temperature would be aected by the overall chamber initial conditions. In order to monitor temperatures at the coolers inlet and outlet, thermocouples were used. However, because of their size, they could not be integrated directly into the wafer. The thermocouples were integrated into the uid lines within one and a half inches from the inlet and outlet of the wafer ports by a small plexiglass connector. The whole system was then placed in the centre of the polystyrene Styrofoam chamber. Centred between the electromagnet poles with the at surface vertical, the microcooler was insulated with half-inch thick pieces of Styrofoam to emulate an adiabatic condition. 4.3.2. Eects of channel layers on the cooling. For the integration of the cooler as described previously, three dierent methods to integrate a channel wafer with the Gd5 Si2 Ge2 block were used. The rst method utilized a crystal bonding of the channel wafer with the Gd5 Si2 Ge2 block. However, the cooler had a leakage immediately when it was placed in the testing chamber. This may be due to the shrinkage of the crystal bond material by rapid thermal change. The second set of tests was conducted on the anodically bonded stack. Beginning at a temperature of about 262 K, the testing was performed and results are presented in Figure 9. The plot of the change in temperature over time shows the response of the sensors for liquid owing into a Si-to-glass wafer combination. Ti and To are the inlet and outlet temperatures, respectively. The electromagnet was turned on and it reached the maximum magnetic eld. After a duration of about 8 s, the electromagnet was turned o. This demagnetization increased the entropy of Gd5 Si2 Ge2 and the uid contacting the Gd5 Si2 Ge2 block was cooled. The initial inlet temperature of 264.8 K was higher than the outlet temperature of 261.3 K. Thus, when the magnet was turned o, the Gd5 Si2 Ge2 block absorbed heat from the uid of the inlet side. After the temperature of the inlet was similar to the outlet temperature, the decrease of the inlet temperature was stopped. The temperature of the inlet continuously increased after 12 s, because Gd5 Si2 Ge2 was not a continuous heat source, but rather an instantaneous source. Also, the temperature of the inlet port increased after 12 s because the cooled inlet uid drove heat ow from the outside uid of the chamber. The outlet temperature also slightly increased, despite an increase of the inlet temperature. The dierence of inlet and outlet temperatures, shown in Figure 9, shows the cooling capability of the MCE cooler. An additional cooler test
Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

726

S. KIM ET AL.

Figure 9. Cooling test of the cooler with the Si-to-glass bonded channel wafer at initial temperature of 262 K.

Figure 10. Cooling test of the cooler with the Si-to-Si bonded channel wafer at initial temperature of 272 K.

was conducted with the Si-to-Si wafer combination, obtained through fusion bonding. The same procedure was followed as the previous cooler tests and Figure 10 illustrates the result. Although the fusion bonding did not hold for long, and although leaking occurred between the wafers, the results were similar to the results obtained from the anodic bonding wafer. The results were signicantly better than results obtained with the glass-to-Si wafer combination, even if the initial temperature was only 272 K. This occurred due to the fact that glass has a lower thermal conductivity than Si. Better cooling was observed with the layer of higher thermal conductivity.
Copyright # 2006 John Wiley & Sons, Ltd. Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er

MAGNETOCALORIC MICROCOOLER

727

5. CONCLUSIONS A magnetocaloric cooler with microuidic channels in Si was designed, fabricated and assembled. The magnetocaloric cooling element Gd5 Si2 Ge2 was characterized for magnetocaloric entropy change to check the feasibility of the cooler, and was found to have the highest entropy change at 262 K. A maximum temperature change of 7 K of the Gd5 Si2 Ge2 block was observed at an initial temperature of 258 K around the highest entropy change peak. The channel wafer bonded to the Gd5 Si2 Ge2 block by crystal bonding. In addition, glass and Si wafers were used as intermediate layers for assembly without leaking at the assembled interface. The cooling test showed that the microcooler had the capability to absorb the heat and cool the uid. The experimental results for thermal characterization of the MCE material and cooler suggest the feasibility of extending the cooler to a magnetic micro-refrigerator.

ACKNOWLEDGEMENTS

This research was funded by NASA under the grant Hydrogen Research at Florida Universities (HYRES) through the support of Clean Energy Research Center. The authors would like to thank Larry Jones at Materials Preparation Center AMES Laboratory, Iowa for their assistance in fabrication and characterization of the alloy.

REFERENCES Bohigas X, Molins E, Roig A, Tejada J, Zhang X. 2000. Room-temperature magnetic refrigerator using permanent magnets. IEEE Transactions on Magnetics 36(3):538544. Kim S, Bethala B, Ghirlanda S, Sambandam SN, Bhansali S. 2005. Design and fabrication of a magnetocaloric microcooler. Proceedings of ASME IMECE IMECE2005-82720, Orlando, Florida, 511 November. Oh KW, Han A, Bhansali S, Ahn CH. 2002. A low-temperature bonding technique using spin-on uorocarbon polymers to assemble Microsystems. Journal of Micromechanics and Microengineering 12:187191. Pecharsky VK, Gschneidner Jr KA. 1999a. Magnetocaloric eect and magnetic refrigeration. Journal of Magnetism and Magnetic Materials 200:4456. Pecharsky VK, Gschneidner Jr KA. 1999b. Magnetocaloric eect from indirect measurements: magnetization and heat capacity. Journal of Applied Physics 86:565575. Pekola JP, Manninen AJ, Leivo MM, Arutyunov K, Suoknuuti JK, Suppula TI, Collaudin B. 2000. Microrefrigeration by quasiparticle tunnelling in NIS and SIS Junctions. Physica B 280:485490. Shirron P, Canavan E, DiPirro M, Jackson M, King T, Panek J, Tuttle J. 2002. A compact high-performance continuous magnetic refrigerator for space missions. Cryogenics 41:789795. Rahman MM, Rosario L. 2004. Thermodynamic analysis of magnetic refrigerators. Proceedings of ASME IMECE IMECE2004-61369, Anaheim, California, 1319 November. Zeng G, Fan X, LaBounty C, Bowers JE, Croke E, Christoerson J, Vashaee D, Zhang Y, Shakouri A. 2001. High cooling power density of SiGe/Si superlattice microcoolers. Materials Research Society Symposium and Proceedings 691:147152.

Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Energy Res. 2007; 31:717727 DOI: 10.1002/er