Thermosiphon Cooling System for a Methanation Reactor: An Experimental Assessment

The passive operation of chemical reactors has been proposed for applications ranging from high flux isotope reactor [1] to natural gas storage system [2]. However, the selection between a single-phase and two-phase thermosiphon depends on the purpose. A single-phase thermosiphon can be used to transport suspended matter in a polymerase chain reactor [3]. However, a two-phase flow benefits from higher capacity to transport heat through the latent heat and, depending on the conditions (geometry, heat flux, flowrate, fluid, etc.), has higher heat transfer coefficients. By using passive cooling compared to active cooling to operate, a cooling system has generated interest in certain fields such as nuclear energy, following the Fukushima incident [4]. With regard to chemical reactions such as methanation, the concept has been proposed previously for $CO2$ methanation itself [5]. The combination of a highly exothermic reaction (Reaction (1)) and a hot spot where the majority of the heat is generated provides an ideal situation for applying passive cooling in the form of a two-phase thermosiphon. In the application, the reactor will be coupled to a solid-oxide electrolyzer that will use the generated steam to produce $H2$ injected in the reactor. Consequently, the steam generation process must be stable.

The biogas methanation reactor presented by Dannesboe et al. [6] is designed as a steam drum with multiple tubes in a single shell. They were able to reach a water pressure of 65 bar (⁠$280∘C$⁠) without the need of a recirculating pump. For our reactor at the laboratory of the Group of Energy Materials (GEM), the reactor was designed to increase the chance of success at stabilizing the steam generation for a single-pass methanation reactor. The reactor’s shell does not work as a steam accumulator; a net upward flow of liquid water improves rewetting and a two-phase mixture leaves the shell.

The cooling system can be operated in thermosiphon mode operation where the recirculating flowrate is dictated by the difference in gravitational forces between the liquid column (downcomer) and the evaporating column (riser) as presented in Fig. 1. In Fig. 1(a), the traditional closed-loop thermosiphon with the evaporator located under the condenser is presented. A thermosiphon operates based on the density difference between the downcomer in which liquid is flowing downward to the evaporator and the riser in which vapor (or a two-phase mixture) is transported upward to the condenser. Far from the critical point, vapor has a much lower density than its liquid form, and thus, a gravitational force difference is generated between the riser and downcomer, which induces a net flow. In a closed-loop thermosiphon, the system’s pressure is dictated by the filling ratio and by the condenser’s capacity. In the case of a co-generation reactor, an open-loop thermosiphon (Fig. 1(b)) is required if the cooling liquid consists of the evaporating steam and the latter is one of the end products of the system. A closed-loop thermosiphon with steam generation in the condenser could be used with a secondary two-phase fluid selected for the temperature range of the reactor’s cooling system; the fluid can be water at a higher pressure. Nonetheless, a second temperature drop would be required, and the added components would generate extra thermal losses and cost.

The configuration used in this analysis is of an open-loop thermosiphon, similar to naturally circulating boiler [7], with steam separation, and the gravitational force is proportional to $(ρl−ρv)gLeff$⁠, where the $ρl$ and $ρv$ are the liquid and vapor densities, respectively, $g$ is gravity, and $Leff$ is the effective distance between the top of the liquid level in the separator and the location in the riser/evaporator where it is equivalent to a sharp transition from 100% water to 100% steam. This net gravitational head is countered by the total frictional losses in the entire loop and by the evaporation pressure drop, and equilibrium is reached when they equate to one another. As the system is open, the back-pressure controller at the steam outlet sets the saturation pressure in the system.

In the previous work [8], a parametric study was completed on the effects of reactants flowrates, reactants pressure, and cooling water pressure on the conversion during $CO2$ methanation. During the experiment, active cooling was performed using a recirculation pump, sending $200−220g/min$ of water close to its saturation condition to the cooling system of the reactor. The pump has a clearance that still allows a small flow to recirculate even if the pump is not activated. Yet, the pump is used to increase the flowrate compared to the passive operation. The use of the recirculating pump was for a few reasons: (1) maintain the flowrate of recirculating water mostly constant regardless of the heat generation and evaporation process, (2) ensure sufficient water recirculation for the range of reactants flowrates, and (3) stabilize the evaporation process for the range of reactants flowrates. The isentropic work required for a pump to recirculate these flows is less than 1 W for a pump head of 300 mbar; the main power consumption comes from the pump’s control system. Therefore, the primary advantage of passive cooling in this context is the increased reliability as a component that is susceptible to failure is avoided. In addition, the use of a recirculating pump also limits the cooling water pressure due to the operating temperature limit of the pump. The operability of the system in thermosiphon operation would allow reaching a higher recirculating temperature, which would be beneficial to the conversion rate. Therefore, the aims of the analysis are as follows: (1) successfully operate the system in thermosiphon mode, (2) vary both the cooling system pressure and the recirculating flowrate to identify if these variables induce or increase instabilities, (3) compare the stability of the system with the active cooling case, and (4) conclude if the thermosiphon operation is a viable operating mode. The comparison will be completed on the bases of the temperature profile in the reactor, outlet vapor pressure, and the response in vapor flowrates leaving the system.

The system published in Ref. [8] is outlined in Sec. 2 with the basis of the analysis of the results. The system's response under thermosiphon operation is presented and discussed in Sec. 3 with the comparisons with the active cooling while varying the pressure (Sec. 3.1) and the recirculating water flowrate (Sec 3.2). The final remarks complete the article in Sec. 4.

Figure 2 presents the simplified schematic of the cooling system experimental test rig with the general dimensions of the reactor. At the reactants gas side, there is a gas pre-heater, an outlet gas condenser, a water separator, and a back-pressure controller not presented in the schematic. The reactor consists of a fixed-bed reactor divided into two sections by a second injection port to test multiple $CO2$ injections as presented previously [8]. A type-k multipoint thermocouple located on the central axis of the reactor and capable of measuring 20 temperatures is used to measure the axial temperature profile of the reaction as depicted in Fig. 2. The quantity and location of each individual temperature measurement along the central axis was selected to improve the ability to capture the magnitude and the location of the hot spot (maximum temperature in the reactor) compared to a uniform spacing. The hot spot should be located close to the inlet due to (1) the high concentrations of reactant and (2) the relatively high catalytic activity of the catalytic pellets. Hot spots were measured by others within the first 0.25 m of their reactors while also using nickel-based catalyst [6,9,10]. As the bed presented in this study consists of two reactive zones, a second group of measurements is located at the start of the second zone.

The cooling system consists of a recirculating water loop with a gravity-based steam–liquid separator. The make-up water is first pumped through its pre-heater before entering the separator. The liquid in the separator is recirculated passively downward, or a recirculating pump is used to actively recirculate the water. Heating of the recirculating line was performed to maintain the water inlet for the shell within $7∘C$ from the saturation temperature. Before the coolant’s inlet, a control valve is installed to help stabilize the evaporation process and flowrate. Following evaporation, the two-phase mixture leaves the reactor to reach the separator. At the vapor outlet, a manual spring-loaded back-pressure controller is used to build the pressure in the system. A Coriolis type mass flowmeter is placed afterward to measure the steam mass flowrate leaving the system.

The flowrate of 24 NL/min $H2$⁠, with its corresponding $CO2$ flowrate, used for the analysis is selected based on the following information.

The $24NL/min$$H2$⁠, which is roughly 5 kW HHV (higher heating value), is within the range of production capacity of a state-of-the-art 70 cells of $80cm2$ solid-oxide electrolyzer with which the reactor is designed to be coupled. The commercial stack could be operated using a range of current densities, which would lead to varying $H2$ flowrate. In addition, any improvement to the cells’ performance would lead to higher nominal current density and higher $H2$ production. Compared to similar reactors [6,9,10], the gas hour specific velocity (GHSV) of approximately 1600/h is lower due to the low reactivity of the catalyst used for this investigation at the low temperature of the cooling system. More reactive catalyst, such as the one used in Ref. [9], which were produced by the same manufacturer, or higher cooling temperature [6,10] could improve the GHSV while still maintaining the temperature profile rigidly at the inlet of the reactor. During the parametric study in active cooling, at higher reactants flowrates, instabilities in the reaction and in the steam generation were identified through oscillations in the reaction temperatures, outlet’s gas composition, cooling system pressures, water level in the separator, and steam flowrate at the outlet of the system. Due to the interdependencies between the temperature inside the reactor, the reaction rate, the overall thermal resistance, the heat available for extraction, the steam generation, and pressure in the cooling system, instability in a single response can be propagated to the other responses. At lower reactants flowrates, the temperature profile was rigidly pushed toward the inlet, which seems to help stabilize all the responses. Consequently, operating only at the lower end of the reactants flowrates can avoid already identified instabilities that could be exacerbated by only using lower recirculation. However, with an estimated 140–160 W of heat losses for the reactor alone, a minimum reactant’s flowrate is required to ensure significant net steam production from the reaction alone. For $24NL/min$$H2$ and $200∘C$⁠, approximately 770 W of heat can be generated by the reaction. In addition, at the GEM laboratory, the height of the ventilation hood is approximately 2 m, which limits the height at which the separator is installed, and thus, limiting the maximum flowrate of steam that can be recirculated in thermosiphon operation. To ensure sufficient control of the recirculating flowrate, there is an upper limit to the quantity of heat that can be extracted. The vapor fraction and heat flux at the hot spot must still be limited to avoid the critical heat flux that could result in dryout or inverse annular flow at the hot spot of the reactor. In these cases, the heat transfer would drastically diminish at the location of highest heat generation. If the heating area were to interact thermally with only vapor, the cooling system would act as insulation. During the parametric study on $CO2$ methanation performed on the reactor, the steam generation at $24NL/min$$H2$ was between 21.67 and $23.34g/min$ and was affected by the inlet reactants temperature of the reactor. With active recirculation of $200−220g/min$ (⁠$G=4.92−5.41kg/m2s$⁠), the vapor mass fraction at the outlet was below 25%. In the thermosiphon mode, the aim is to limit the vapor mass fraction to be under 50%.

To activate the thermosiphon operation, the recirculation pump and the subsequent valve, presented on the simplified schematic of the cooling system, are shut, while the valve in the direct line is opened. With fixed reactants flowrates, and thus, an almost fixed heat generation, the stability of the system as a function of pressure and recirculating flowrate are assessed.

The temperature profile in the packed bed (and the subsequent conversion) can take significantly longer to stabilize compared to the cooling system’s pressure. In the previously published work, the effects of the water pressure on the conversion and on the temperature profile were identified. Under active cooling with the same reactant flowrate, the conversion increased by less than 1% between 12 and 15 bar G, and thus, the conversion changes between pressure level is negligible on the total steam production. In addition, for the selected reactant flowrate, only a limited change in the captured temperature profile along the central axis was observed within the same water pressure range, as presented in Fig. 4. Therefore, the location of the hot spot and the distribution of the heat generation remain similar. For this investigation, conversion is not measured and the operating points are changed when the cooling system has settled.

As mentioned previously, the analysis will be limited to the $H2$ production of the commercial solid-oxide electrolyzer to be used in the coupling. Increasing the reactant flowrate would lead to higher steam generation and lower vapor fraction at the outlet of the cooling system. To partially capture the effect of higher reactant flowrate, the control valve located at the inlet of the reactor is partially closed to decrease the recirculated flow. The total height between the start of the reaction zone and the maximum water level in the separator is of approximately 1.5 m. However, the effecting height difference, $Leff$⁠, is smaller than this value as the water level is lower and the evaporation is gradual. Yet, the total head due to gravity should be of the order of 100 mbar depending also on the temperature, which affects the densities of both liquid and vapor. Consequently, the pressure drop across the control valve must equate to the gravitation force minus the other frictional pressure drops and the momentum pressure drop in the cooling system. The following are some limitations to this analysis:

• Any significant pressure drop in the recirculation line can help stabilize the evaporation process by augmenting the liquid pressure drop [11].

• Increasing the reactants flowrates would increase the steam generation, which in turn should increase the effective distance between the water and effective vapor column. However, when increasing the reactants flowrates, at some point, the hot spot starts moving away faster from the inlet, which negatively affects the effective distance $Leff$⁠. Therefore, the relationship between recirculating water and the reactants flowrates might have a maximum located within the operating range of the reactor depending on the relative size of the two effects.

• The vapor–liquid separation unit and the subsequent vapor line act as a buffer volume that help stabilize the vapor flowrate and pressure, and thus, the stability assessment is performed not only on the evaporation process.

The operating conditions during the two analyses is presented in Table 1. Changing the reactant flowrate and/or changing the reaction pressure can affect the conversion and heat generation profile as indicated previously [8]. If the hot spot is no longer rigidly at the inlet, the system can face additional sources of instabilities. Consequently, to ease the comparison between the two cooling mode, these variables are maintained constant. The response in vapor flowrate at the outlet of the system is to be compared to the active cooling case through the range over which 98% of measurements over time lie. The mass flowmeter is the one presented previously in Aubin et al. [8] with the 0.5 s damping to remove noise. The various measurement devices used during this investigation along with their uncertainties are presented in Table 2.

As the saturation pressure increases, the saturated liquid density, thermal conductivity, and viscosity decrease while the opposite occurs for the saturated vapor. The decrease in density difference directly impacts the thermosiphon potential. However, the decrease in latent heat and increase in the liquid’s thermal conductivity should increase the $Leff$⁠. On the other hand, the summation of pressure drops across the various restrictions should increase with viscosity and density. In addition, the evaporation process and the flow pattern resulting from it provides the effective column difference. Thus, the difference in gravitational force is dependent on the property already listed in addition to the surface properties. Consequently, the effect of pressure on the recirculating flowrate is multifaceted. The operation of the system at multiple cooling water pressures is presented in Fig. 3(a). For each pressure level, the system was operated until a pseudo-steady state was reached. The transient operation of the system, displayed is the figure as the time between the green rectangles, shows instabilities that could damage the electrolyzer. Consequently, the downstream steam line is required to handle such instabilities.

The variations in pressure measured in passive operation is presented in Fig. 3(b) with a comparison with passive cooling indicating similar size variation in the pressure over time. When comparing the results of both subfigures, the larger variations correspond to points displaying transient behavior. The manual back-pressure controller in the vapor line is sensitive to the flowrate, and thus, if the evaporation process has yet to settle, the pressure continues to drift slightly, which in turn affects the evaporation flowrate. In the final system, the back-pressure controller will be changed to an electro-pneumatic unit with finer control of the pressure. The water level in the steam separator was also maintained as constant as possible to ensure the change in effective column difference is only impacted by the evaporation process. At the start of the operation, the control valve at the inlet of the reactor was set to reach a vapor fraction of 30% at the outlet. A typical response in the recirculating water flowrate and produced vapor flowrate is presented in Fig. 3(c).

The comparison between the passive and active operation of the cooling system is presented in Fig. 3(d) with the average vapor flowrate and 98% of the measurement range. The results indicate that the 98% measurement range (within $±9.82%$ for all passive cooling) is similar to the active cooling (⁠$±9.5%$⁠). The difference in average flowrate measured could be the result of a slight difference in certain operating variables that change the quantity of heat that can be extracted. One such variable is the temperature of the reactive gases sent to the reactor, where during the thermosiphon operation, the reactive gases entered the inlet manifold $70−75∘C$ higher than the active cooling case presented, inflating the quantity of steam generated. As the system could still have some transient behavior, the average steam flowrate at 14 bar G was slightly lower than the other pressures. With the increase in pressure, the latent heat reduces, but within the pressure tested, this gain is only 1.91%. Therefore, the potential gain can be easily hidden by the change in other variables such as the subcooled temperature of the water at the inlet.

Figure 4 presents the reaction temperature profile to indicate the similitude between the active cooling published previously and the passive cooling on the reactor’s temperature profile along the central axis of the reactor. Within the operating range of the reactor, the increase in cooling pressure moved the hot spot toward the inlet marginally. In the previous work, the conversion increases by about 1% within the same total pressure range and, thus, should not affect significantly the total steam generation.

The variation of the cooling system pressure had no significant effect on the reaction’s behavior or the stability of the system. The absence of increased instability with the rise in cooling system pressure suggests that it may be possible to extend the pressure range in passive cooling beyond the validated range in active cooling.

Figure 5(a) presents the recirculating flowrate and steam flowrate, along with the corresponding vapor fraction at the outlet of the evaporator, displayed at the bottom of each green rectangle, assuming negligible condensation/evaporation occurring in the separator. The first operating points consists of recirculating approximately 85.96 g/min while evaporating 20.53 g/min, equating to a 23.9% vapor mass fraction. The heating occurring in the vapor line to avoid condensation is performed manually. This could potentially cause condensation in the line if the heating set point is not adjusted properly and if the back-pressure controller is not capable of effectively controlling two-phase mixtures. The interplay of both issues may have contributed to the behavior observed at approximately 5.8 h. During the event, the valve was also closed to restrict the recirculation to 73.79 g/min; the resulting vapor mass fraction is 28.7%. After reaching a pseudo-steady state, two more operating points were tested: 62.65 g/min (33.7%) and 45.27 g/min (46.5%). The mass velocities of the recirculating flowrates at the four vapor fractions tested are presented in Fig. 5(b) along with their standard deviations. The reduction in the standard deviation percentage in Fig. 5(b) and in the peak-to-peak pulsation amplitude shown in Fig. 5(a) suggests a potential stabilization of the recirculating flowrate. However, this may be influenced by the effect of the restriction on stabilizing the recirculating flowrate and the average recirculating flowrate. To separate the effect of the restriction from the effect of changing the vapor fraction, it is necessary to decrease the latter by increasing the reactants flowrates. It should be noted that there were some instabilities during the operation in active cooling for high reactants flowrates. In active cooling, the steam generation reached 34 g/min for 36 NL/min $H2$⁠, which would result in a minimum vapor fraction of at least 39.5%, assuming the same maximum recirculating flowrate of 85.96 g/min, limiting the flexibility of the operation. In addition, the evaporation process can change significantly between 24 NL/min $H2$ and 36 NL/min $H2$ due to the heat generation profile. At higher flowrates, the hot spot extends over a larger length of the reactor while generating more heat to be extracted. The 770 W of heat that can be theoretically generated by the reaction with 24 NL/min $H2$ becomes 1150 W for 36 NL/min $H2$⁠. Using linear extrapolation between temperature probes on the data in active cooling published previously, an estimated 225 mm is between the two locations (one on the ascent and one in the descent) where the temperature profile crosses the 350 °C level for example with the lower flow rate, compared to 265 mm for the larger flow rate. However, the $H2$ conversion drops from 97.84% to 97.17% between the two flowrates [8], reducing slightly the quantity of heat available for steam production. Finally, the hot spot for the higher flow rare is no longer rigidly at the inlet; in the analysis of the effect of the reactant flowrate published previously, the hot spot was located between the 25- and 50-mm probing locations for the lower flowrate compared to between the 75- and 125-mm locations for the larger flowrate.

Otherwise, as with the effect of water pressure, the higher variations observed in the outlet pressure of certain passive cooling cases tested, as presented in Fig. 5(c), could be attributed to a pseudo-steady state. For the reactants flowrates tested, the average steam flowrate and the 1–99% range (Fig. 5(d)) indicate comparable behaviors within the range of recirculating flowrates, including active cooling. The 1–99% measurement range of all the passive cooling cases are within $±9.74%$ compared to $±9.5%$ for the active cooling. Similar to the case of the effect of evaporating pressure, the thermosiphon operation was operated with a higher inlet reactants temperature, and thus, the system was able to evaporate a slightly larger quantity of steam. On the other hand, the lower quantity generated for the 23.9% case compared to the other passive cooling could be due to the remnant of some transient behavior.

The analysis assesses the viability of operating in thermosiphon mode with reactants flowrates in the proximity of 24 NL/min $H2$ to maintain some of the operation flexibility that active cooling provides, while decreasing the risk of failure and cost of the system.

The exothermicity and kinetics response of $CO2$ methanation makes it an ideal candidate for cooling using a thermosiphon. This study tested a methanation reactor designed to be cooled using partially evaporating water driven by a recirculating pump but in thermosiphon operation instead and compared the system stability (flowrates, pressures, etc.) to the active cooling operation. The comparison was extended to include varied cooling water pressure and vapor fractions at the outlet of the cooling system in the thermosiphon mode to stress the system, and the results indicated a limited impact on the steam flowrates leaving the system and on the pressure in the evaporation process. However, changing the vapor fraction at the outlet of the reactor using a restriction resulted in the stabilization of the recirculating water flowrate. Therefore, the use of a restriction at the inlet of the evaporator limits the ability to assess the effect of the recirculating flowrate on the stability of the pressure and steam flowrate. The restriction can stabilize the recirculating flowrate and the evaporation process independently of the average recirculating flow. Furthermore, the combined volume of the separator and of the steam line before the back-pressure controller may act as a buffer, leading to the limited impact of operating conditions on the steam flowrate leaving the system.

To conclude, the passive operation of the cooling system appears to be a viable option that offers increased safety and could extend the range of cooling water pressure, both by avoiding the use of a recirculating pump. The system was tested with a $H2$ flowrate that was roughly in the middle of the expected flowrate range produced by the electrolyzer that will be connected to the reactor. During passive cooling, the instabilities in the steam flowrate leaving the system, quantified as the range where 1–99% of the measurements lie, were limited (within $±9.82%$⁠) at this $H2$ flowrate, comparable to the instabilities during active cooling (⁠$+/−9.5%$⁠). Future work will involve investigating regions of operation, mostly by changing the reactants flowrates, which could potentially result in instabilities and exploring methods, such as using a restriction at the inlet of the shell, to stabilize the system’s response. As the instabilities did not increase when increasing the cooling water pressure, while in gravity-driven operation, water pressures above the temperature limit of the recirculating pump will be further investigated.

This research was funded by the HOTCAT4STEAM (I.D. SI/501825) project from the Swiss Federal Energy Office, by the European Union’s Horizon 2020 under grant agreements n. 731224 (BALANCE, topic LCE-33-2016), by GAZNAT in the scope of a project on the development of a SOE-methanation coupling, and by the EPFL-Wallis and HES-SO joint Valais demonstrator. The authors would like to thank Hitachi Zosen for supplying the catalyst.

There are no conflicts of interest.

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.