Figure 1. Pressurizing to 8.4 psig, cooling 4 min,
followed by a heat/cool cycle
Vish V. Viswanathan, Robert Wegeng and Kevin Drost
This research focused on investigating use of Pacific Northwest National Laboratory's (PNNL) microchannel architecture for adsorption. For our study, we chose zeolite adsorbent for thermal swing adsorption of CO2. Single- and multi-cell adsorption devices were fabricated. While the multi-cell devices could not be used for testing due to unresolved cross-talk issues, a solid groundwork was laid for further development of such devices. Thermal swing adsorption was done on a single-cell device. The response time of the device was determined and compared with commercially existing technology. The approach used in this work can also be used for adsorption of ammonia into carbon.
It was assumed that the CO2 that reaches the zeolite micropores is instantaneously adsorbed. Hence the time required for CO2 to reach the zeolite micropores was assumed to control the mass transfer kinetics. Ignoring Knudsen diffusion, it was estimated that 95% of CO2 reaches the zeolite particles in 30 seconds, based on semi-infinite diffusion. Once at the surface of the zeolite beads, it was estimated that CO2 would diffuse inside the 10 Å micropores within 0.1 seconds. This analysis indicated that heat transfer kinetics would be the rate-limiting step for thermal swing adsorption.
These devices consisted of microchannel cavities to house the adsorbent. Two different designs were used to provide heat exchange to the cavity. It was anticipated that one design would respond faster than the other design to temperature changes in the heat exchange fluid. The individual shims for the heating and cooling channels were bonded by diffusion bonding. However, this led to cross-talk between the gas and liquid channels because the relatively high pressure distorted the shims on either side of the channels, depressing them slightly into the water channels below and preventing the depressed area from bonding to the shim above. This caused heat exchange liquid to leak from the header area to the gas side and vice-versa.
The single-channel device consisted of a copper block with copper tube welded to it for cold and hot water flow. Water flow was maintained at 200-260 mL/min. The temperature of the cold water stream was 4°C; the hot water temperature varied between 60-80°C. Prior to adsorption, CO2 was fed into the device and allowed to flow through for five minutes at 658 mL/min (1.16 g CO2/min) with the device at ambient temperature. The capacity of zeolite ranged from 0.1-0.26 g CO2 per g zeolite over the temperature and pressure range of this work. Hence, this flow rate was considered sufficient to supply the required amount of CO2 to the device. The gate valve downstream of the adsorber was closed so the pressure inside the device was ~12 psig. The system was allowed to stabilize for five minutes to ensure that the adsorbed CO2 in the zeolite had reached equilibrium with the CO2 in the gas phase. The three-way valve upstream of the device was then turned so the gas supply was cut off. The device was cooled for 10 minutes, then heated for 10 minutes. The CO2 pressure was recorded with the change in temperature of the device. The above procedure was repeated for various initial temperatures of the device.
The pressure reached a plateau within two minutes of heating or cooling the device. The rate of increase of pressure was significantly higher than the estimated rate assuming pressure change due to ideal gas law. Figure 1 shows the results of a run in which the initial pressure was 8.4 psig. Data on commercial adsorbents indicated that the adsorption capacity of zeolite at -78°C and 10 torr is 27%, while it was 22% by weight at 5 psig and 4°C. Hence the results of using a wider temperature range (-93°C for adsorption and ~ 60°C during desorption) under Martian ambient conditions at night should be quite promising. Optimizing the device for improved heat transfer to the zeolite while minimizing heat loss should allow faster response of the system to temperature change. The faster pressure increase during heating was indicative of the contribution of desorption, while the more rapid pressure drop during cooling was indicative of adsorption.
Figure 1. Pressurizing to 8.4 psig, cooling 4 min,
followed by a heat/cool cycle
Figure 2 shows the hysteresis in the pressure versus temperature plot during adsorption and desorption. The zeolite temperature was expected to lag that of the copper block surface where the thermocouple is placed. Hence, during adsorption (cooling), the temperature of the zeolite was expected to be higher than the copper block surface. This would correspond to a higher CO2 pressure. During desorption, the zeolite temperature would be lower than the copper block surface, which corresponds to a lower CO2 pressure. So for a fixed pressure, half of the difference in temperature for the adsorption and desorption curve would correspond roughly to the temperature lag of the zeolite. As expected, this lag was maximum at the mid-range of temperature. At the extreme temperature values, which correspond to the beginning or end of the cooling or heating cycle, the zeolite has had sufficient time to catch up with the temperature of the copper block. The maximum temperature lag was about 10ºC. Because the copper block temperature changed at the rate of about 20°C/min, this corresponds to a lag of 30 seconds during the middle of the heating or cooling half cycle. For a device unoptimized for fast heat transfer, this was quite fast. By allowing faster change in the copper block temperature, and minimizing zeolite temperature lag (thinner zeolite shims, low heat loss), the temperature lag may be minimized to ~2-3 seconds. Hence this would allow cycling the device almost as fast as the copper (or any other metal) shim temperature is changed. A half-cycle time of 10 seconds is expected to be achievable.
Figure 2. Hysteresis curve during adsorption/ desorption
Discussion with vendors of zeolite adsorbent indicated that the half-cycle time for temperature swing adsorption currently is 1.5-2 hours. The adsorbent beds range from 55 gallons to tanks that are 10 feet in diameter by 10 feet tall. Clearly, our device is much more suited for fast cycling, thus allowing use of smaller, lighter devices.
In the future, we plan to focus on repeating the above experiments for ammonia adsorption. The device design will be optimized for improved heat transfer. A multichannel design will be developed after the above tasks are completed.
The authors would like to thank Rick Cameron, Dean Matson, and Don Stewart for designing and fabricating the adsorption device.
|