The objective of this study was to develop a strategy for miniaturizing heat exchangers (HXs) used for the thermal management of sorbent beds within adsorption refrigeration systems. The thermal mass of the microchannel heat exchanger (MCHX) designed and fabricated in this study is compared with that of commercially available tube-and-fin HXs. Efforts are made to quantify the overall effects of miniaturization on system coefficient of performance (COP) and specific cooling power (SCP). A thermal model for predicting the cycle time for desorption is developed, and experiments are used to quantify the effect of the intensified HX on overall system performance.
Adsorption refrigeration systems have several advantages over mechanical vapor compression cooling systems including no greenhouse gas refrigerants, the ability to use waste heat, and fewer moving parts. The main problem with adsorption refrigeration systems is bulkiness tied mainly to the low energy efficiencies inside the sorbent bed. Prior research has been performed to improve the energy efficiency of the thermal management technology within sorbent beds [1–3]. Higher heat exchanger (HX) efficiencies make it possible to reduce the size and weight of adsorption chillers, since the heat and mass transfer rates and thermal mass of sorbent beds strongly affect system cycle times . No prior efforts have been made to use microchannel technology to reduce the size of thermal management heat exchangers within sorbent beds.
In this paper, we design and fabricate a microchannel heat exchanger (MCHX) that is used to reduce the thermal mass and cycle times of a sorbent bed. This includes a header design which improves the coefficient of performance (COP) and specific cooling power (SCP) of the overall system. Further, we compare the thermal performance of the microchannel heat exchanger with that of a conventional tube-and-fin heat exchanger. Heat transfer experiments are compared with modeling results.
Heat Exchanger Design
where cAds is the specific heat of the sorbent (J/kg °C), mAdHX is the mass of the heat exchanger (kg), cAdHX is the specific heat of the metal (J/kg °C), Ha is the heat of adsorption (J/kg sorbate), and T3 and T1 represent the temperature (°C) at the end of desorption and adsorption, respectively.
where τcyc is the cycle time for the adsorption/desorption cycle (s). The larger the SCP, the less mass that the system entails.
Sorbent Bed Considerations.
In the design of a sorbent bed, key factors affecting system performance include the mass transfer of the sorbate through the bed (interparticle flow) and sorbent (intraparticle diffusion) as well as the heat transfer through the bed . Cacciola et al.  studied the effect of heat exchanger designs on the performance of adsorption heat pumps and found that increasing the metal mass, to enhance heat transfer rates within the sorbent bed, actually reduced both the COP and SCP of the system. Less metal mass in the bed reduces the mass of the system while reducing the time needed to heat up the bed yielding an increase in SCP. Less metal mass also reduces the amount of energy needed to run the cycle which translates into a higher COP. A tradeoff exists between the amount of metal to be added to increase the conductivity of a sorbent bed and the need to minimize the thermal mass of the sorbent bed [5,6].
Table 1 summarizes adsorption heat exchanger designs reported to date. Freni et al.  developed both coated and granular-type adsorption heat exchangers. The coated heat exchanger had a metal-to-sorbent mass ratio of six, compared to a ratio of 1.96 for the granular-type exchanger which was reported to have a higher COP. Further, Chang et al.  designed a granular and flat-tube type heat exchanger with corrugated fins for sorbent beds in order to increase the heat transfer area and reduce the thermal transport distance within the bed. Both Freni et al. and Chang et al. used conventional extruded tubing to make the adsorption heat exchanger. Sharafian and Bahrami compared existing sorbent bed designs and showed no bed designs that have been experimentally confirmed to have a metal-to-sorbent mass ratio less than one .
Heat Exchanger Design.
The overall concept of the adsorption heat exchanger developed in this paper is shown in Fig. 1. The unit cell consists of two microchannel heat transfer plates, a corrugated fin structure between the plates, and two end plates for attaching headers. The 3003 aluminum structure was brazed together followed by adding granular sorbent in between the fins. The powder was constrained with the use of a metal mesh wrapped around the structure between the header plates.
To eliminate heat transfer bottlenecks within the bed, thermal resistances within the bed were analyzed as shown in Fig. 1, where R1 is the convective thermal resistance from the heat transfer fluid to the channel wall; R2 is the conductive thermal resistance through the microchannel wall; R3 is the thermal contact resistance between sorbent grains and the exterior channel surface; Rf is the thermal contact resistance through the fin; R4 is the thermal contact resistance between sorbent grains and the fin; and R5 and R6 are the conductive thermal resistance through the sorbent bed parallel to the fins and between the fins, respectively.
where Nu is the Nusselt number, k is the thermal conductivity of the heat transfer fluid, and Dh is the hydraulic diameter of the microchannel. A Nusselt number value of 8.01 was used in this work assuming a fully developed laminar flow in the rectangular microchannel at constant wall heat flux. The Nusselt number was determined by Kandlikar . The convective heat transfer coefficient was found to be 1080 W/m2 K.
Below, Table 2 summarizes the calculations associated with the thermal transport within the microchannel heat exchanger. The remaining thermal resistances were determined by the dimensions of the heat exchanger, the thermal conductivity of the aluminum and silica gel, and the thermal contact resistance between the aluminum and the silica gel. The thermal contact resistance between silica gel and aluminum was taken to be 2.0 m2 K/kW based on the work of Sharafian et al.  and Zhu and Wang  who investigated the contact resistance between different granular sorbents and copper to range between 0.83 and 2.55 m2 K/kW. The effective thermal conductivity of the silica gel powder was made to be 0.147 W/m K based on averaging the effective thermal conductivity from the literature as summarized in the work of Freni et al.  and Sharafian et al. .
To date, there have been no reports of using microchannels in adsorption heat exchangers. This is likely due to the thermal transport within the heat exchanger being normally governed by conduction through the sorbent bed rather than by convective transport distances. With shorter fin spacing in the bed, the conductive thermal resistance of the sorbent between the fins (R6) can be significantly reduced. The above analysis suggests that heat will flow from the bed to the fins, down the fins, and into the fluid flowing through the microchannel, and that the convective thermal resistance within the heat exchanger can become the rate limiting step.
The use of microchannels was pursued to minimize convective thermal resistance and reduce the overall mass of the heat exchanger. Figure 2 shows the design of a lamina with 500 μm deep microchannels used in the current study (dimensions in mm). Header mass was reduced by focusing the ends of the microchannels (Fig. 3) compared with tube-and-fin designs. The heat exchanger lamina was produced using photochemical machining to enable flow features at the channel entrance and exit to help distribute flow.
Lamina thickness was minimized by using posts in the flow channel to reduce spans and consequent deflections during operation. The region marked by the dashed line in Fig. 4 represents the largest span. To model deflections, the plate was considered flat and rectangular with fixed edges, subjected to a uniformly distributed load. Based on plate mechanics, the allowable pressure that the device can operate with acceptable deflection (5% of channel height ) was found to be 
where σY is the yield strength of aluminum, b is the span, and tl is the thickness of the top lamina. Based on an allowable deflection of 55 μm, the allowable pressure was found to be 70.8 psi which is significantly higher than the expected working pressure of 9.8 psi based on pressure drop estimates suggesting that additional mass savings are possible.
Experimental Plan and Modeling
To compare the thermal mass ratio of the two module designs, extruded-tube-and-fin and microchannel heat exchangers were designed to cool 410 g of silica gel. Figure 5 shows the cross section of the extruded tube used in this study. The heat exchangers both consisted of five unit cells containing 82 g of silica gel with identical bed heights. A summary of the dimensions for the extruded tube and microchannel is provided in Table 3. The length and width of the microchannels were chosen so the volume of the unit cell beds would match. As shown in Table 4, the microchannel design shows a 19.2% reduction in thermal mass over the tube-and-fin design. Most of this weight savings is in the channels and headers. The metal-to-sorbent mass ratio of the microchannel unit cell was found to be 66.3%, a 63.6% reduction over the lowest metal-to-sorbent mass ratio reported to date.
To evaluate the thermal performance of the heat exchanger designs, microchannel and extruded-tube-and-fin unit cells were fabricated each capable of containing 82 g of silica gel. Figure 6 shows the two unit cells. Microchannel laminae were patterned by photochemical machining. An unpatterned top lamina was laser welded to the patterned bottom microchannel lamina to form the microchannel plate. Two microchannel plates, a section of corrugated fins, and two header plates were assembled using vacuum brazing at around 600 °C.
A comparison of braze joints produced using brazing paste versus foils is shown in Fig. 7. It was found that the brazing foils provided larger fillets at the joints without voids. An interferometric microscope was used to the image the fillets formed by the foils (Fig. 8), showing the fillets to be 0.79 mm wide by 0.73 mm tall. Brazing allowances of 75 μm were found to cause discontinuous fillets with voids. On the other hand, allowances of 25–37.5 μm made the exchangers hard to assemble. Therefore, a clearance of 50 μm was used.
Figure 9 shows the setup for vacuum brazing in the furnace. H, M, and L thermocouple locations represent the high, medium, and low temperature within the brazed workpieces. Magnesium chips were used as an oxygen getter. The brazing cycle is shown in Fig. 10. During the cycle, early pressure spikes are due to moisture or binder burnout. At around 300 °C, the magnesium chips sublimate to react with any remaining oxygen, water vapor, and surface oxides.
After brazing, headers were joined to the brazed assembly by gas tungsten arc welding. The fabrication procedure for the tube-and-fin module was the same as for the microchannel module except that it used off-the-shelf tubing. Once the two modules were fabricated, a metal mesh to contain the powder was attached onto one channel by spot welding. Then, the powder was inserted between the fins, and the metal mesh was wrapped around the unit cell and spot welded (Fig. 11 bottom).
Silica Gel Adsorption Test.
Prior to running thermal performance tests, experiments were conducted to determine whether the latent heat of adsorption of the silica gel under ambient conditions would impact thermal performance. This was a concern because the open bed areas of the two unit cells were different. Here, open bed area is defined as the bed height times the bed length. The microchannel unit cell had a 188:1 open bed area-to-thickness ratio, while the tube-and-fin unit cell has a ratio of 370:1.
To simulate the bed thickness of the two unit cells, two beakers were filled with silica gel to one half of the thickness of each bed. Before beginning the adsorption tests, the silica gel was desorbed in a controlled atmosphere at high temperature for 1 h and allowed to cool to room temperature. Afterward, the beakers were removed from the oven and weighed every 3 min to determine the rate at which the silica gel would adsorb atmospheric gases under ambient conditions. It was determined that for a 3-min adsorption step, the unit cells would adsorb 0.18 g and 0.11 g per 100 g of silica gel for the tube-and-fin and microchannel unit cells, respectively. These values were considered conservative, since the actual thermal performance experiments were to be started at 130 °C where the adsorption rate would be lower.
Accordingly, the maximum heat of adsorption for the thermal performance experiments was determined to be 0.35 kJ for the tube-and-fin unit cell and 0.214 kJ for the microchannel unit cell. These values are 5.5% and 3.4%, respectively, of the sensible heat to be extracted from the sorbent bed during adsorption which was considered acceptable.
Module Thermal Performance Testing.
To assess the thermal performance of the two unit cells, they were subjected to thermal cycles to simulate the adsorption step in an adsorption-cycle heat pump. The two beds were heated to 130 °C and cooled to 50 °C using a heat transfer fluid. Figure 12 shows the instrumenting and packaging of the tube-and-fin unit cell. For each unit cell, the bed temperatures were collected at multiple locations as a function of time at three different flow rates.
Figure 12 shows the experimental apparatus for measuring thermal performance. It includes two fluid reservoirs to store the cooling fluid, Paratherm NF. Nitrogen was applied to the head space of the reservoirs to drive the fluid flow. The fluid was measured at the inlet and outlet for both temperature and pressure. The flow rate was measured using catch-and-weigh methods as shown in the picture. In order to monitor the cooling of the bed with time, six T-type thermocouples were inserted into the sorption bed in flow-wise locations (t/c 1–6). All data were collected using a National Instrument Compact data acquisition system. Insulation was wrapped around the beds to minimize thermal losses.
An ANSYS transient thermal model was used to simulate the adsorption tests and investigate the effect of the microchannels on heat transfer. The thermal models were developed to simulate the heat transfer at thermocouple 1, since the fluid inlet temperature can be estimated with good accuracy. Based on the hydrodynamic and thermal entry length in Table 5, it was found that the flow did not reach the fully developed regions until 140 mm and 180 mm for the microchannel and tube-and-fin heat exchangers, respectively. Therefore, the Nusselt numbers of the channels were obtained assuming thermally developing laminar flow in a rectangular microchannel  and a circular tube  at constant wall heat flux
where Nu is the Nusselt number from Shah and London, Re is the Reynolds number, Pr is the Prandtl number, D is the hydraulic diameter, and L is the flow length. Therefore, the heat transfer coefficients were calculated to be 1190 W/m2 K for the microchannel unit cell and 908 W/m2 K for the tube-and-fin unit cell.
Figure 13 shows the geometry and mesh (hexahedron) used in the computational models. The figure shows half tubing on the top and bottom with corrugated fins in the vertical direction situated between the silica gel bed. Due to geometric symmetry, only two corrugated fins, one sorbent bed, and two half sorbent beds are modeled. The boundary conditions are the convective heat transfer coefficients on the top and bottom with thermal insulation on the sides due to geometric symmetry.
Results and Discussion
The cooling time as a function of thermocouple location of both unit cells is shown in Fig. 14. As expected, for both unit cells, the time needed for the bed to be cooled to 50 °C decreases as the volumetric flow rate of Paratherm NF increases and as the distance from the channel inlet decreases. The percent reduction in cooling time for the microchannel unit cell compared to the tube-and-fin heat exchanger is shown in more detail in Table 6.
The results of the experiments show on average a 9.2% decrease in the cooling time for the microchannel unit compared with the tube-and-fin unit. The improvement is less at higher flow rates, which is likely attributed to the characteristic change of the flow regimes within the entire flow length. Given the significant longer flow length for the tube-and-fin design, the portion of flow length as thermally developing became more pronounced as flow rate was increasing. This led to the shrinking advantage of the microchannel as shown in Table 6. The design of the microchannel plate to enable a larger portion of the flow to be fully developed would further improve the percent reduction in cooling time. Thermocouples 3 and 6 in the microchannel unit show longer cooling times than expected at all volumetric flow rates. This was likely attributed to worse effective thermal conductivity in the bed or high contact resistance between the thermocouple and the bed, which was likely caused by poor packing density of the sorbent leading to more air pockets in those measurement locations.
Figure 15 shows a comparison of the modeling results with the actual performance of the heat exchangers. The significance of these results is that the models predict a fairly small improvement in cooling time for the microchannel unit. Table 5 provides some insights about the channel geometry in term of hydraulic diameter, which had only modest (30%) contributions to improving the convective heat transfer coefficient for the microchannel heat exchanger. The fact that the flow was thermally developing for both cases (but at slightly different amounts) also had some effect on the thermal enhancement. In addition, the fins inside the tubes also helped overall heat transfer given their high fin efficiency. Based on these findings, further design optimization could further improve the performance of the microchannel design.
Even though the model captured the trend, there are still significant differences between the results of the experiments and the thermal models; 33.8% error for the microchannel heat exchanger and 25.4% error for the tube-and-fin heat exchanger. This difference is mainly attributed to the uncertainty associated with the effective thermal conductivity of silica gel which was estimated based on values from the literature. Factors such as particle size and the packing density could have significantly affected the effective thermal conductivity. From the summarized thermal conductivity packed bed of dry silica gel reported in the literature , the values range from 0.106 to 0.198 W/m K. Comparing the reported values with the expected effective thermal conductivity of 0.147 W/m K, a maximum discrepancy of 27.8% can be attributed.
Second, the models have less thermal mass than the experimental modules due to the headers which were not modeled for simplicity. Therefore, it is expected that the model would cool faster than the experiments. After taking into account the additional thermal mass associated with the experiment, the increased heat capacity in the model could be as much as 9.0% and 15.4% for the microchannel and tube-and-fin heat exchangers, respectively. This can increase the model cooling time of the bed proportionally.
Predicted COP and SCP of Unit Cell.
Based on these results, the COP and SCP values of a basic adsorption chiller are presented in Table 7 for both the microchannel and tube-and-fin heat exchangers. The calculation is based on the following conditions: a working capacity of 7% for the sorbent, bed temperature ranges between 50 and 130 °C, an evaporator temperature of 12 °C, and a condenser temperature of 60 °C. In addition, the microchannel module is expected to have a 9.8% shorter cycle time than the tube-and-fin module based on the results of the thermal testing. These results show a potential 24% increase in the SCP through the use of a microchannel heat exchanger.
A granular-type microchannel adsorption heat exchanger for integration with a sorbent bed was developed in this study. It was found to reduce the thermal mass ratio by almost 20% over a conventional tube-and-fin heat exchanger. Further, the microchannel heat exchanger was found to reduce cooling time by nearly 10% due to a higher convective heat transfer rate within the microchannel. Based on a simulation model and a thermal resistances analysis, this is the result of the heat transfer bottleneck being within the flow channel, not within the sorbent bed. Under these conditions, the microchannel adsorption heat exchanger could increase the predicted coefficient of performance and specific cooling power of an adsorption refrigeration system by 4% and 24%, respectively, compared with that of a tube-and-fin heat exchanger.
This work was performed for the U.S. Department of Energy under Contract No. DE-AR0000369 (CFDA No. 81.135) and funded by the U.S. Department of the Navy in conjunction with the U.S. Department of Energy (DOE) Advanced Research Projects Agency-Energy (ARPA-E). The authors would like to thank Patrick McNeff and Dr. Ki-Joong Kim for their assistance with laser welding and braze paste application.
Advanced Research Projects Agency (Grant No. DE-AR0000369).
- b =
dimension of the short side of plate
- c =
- COP =
coefficient of performance
- hc =
convective heat transfer coefficient
- hf =
- Ha =
heat of adsorption
- Hv =
latent heat of vaporization of refrigerant
- k =
thermal conductivity of heat transfer fluid
- Kal =
aluminum thermal conductivity
- KAds =
thermal conductivity of silica gel
- lf =
distance between fins
- m =
- M =
- Pmax =
maximum allowable operating pressure
- QDes =
heat required for the desorption process to occur
- QEv =
amount of chill produced in the evaporator
- QPH =
heat of preheating
- Rcont =
thermal contact resistance between aluminum and silica gel
- Rf =
conductive thermal resistance through the fin
- R1 =
convective thermal resistance from heat transfer fluid to channel wall
- R2 =
conductive thermal resistance through wall
- R3 =
thermal contact resistance between sorbent grains and microchannel surface
- R4 =
thermal contact resistance between sorbent grains and fin
- R5 =
conductive thermal resistance through sorbent bed parallel to the fins
- R6 =
conductive thermal resistance through sorbent bed between the fins
- SCP =
specific cooling power
- t =
- Tcon =
- Tev =
- T1 =
end of desorption
- T3 =
end of adsorption
- w =
width of microchannel
- α =
ratio of short side to long side of the microchannel
- Δxcool =
change in the differential refrigerant mass uptake during the precooling process
- σY =
yield strength of aluminum
- τcyc =