Hydrogen Production from Natural Gas Using Hot Blast Furnace Slag: Techno-economic Analysis and CFD Modeling

Abstract

A process for thermal decomposition of methane to hydrogen and solid carbon is presented and examined. It utilizes the high-temperature heat from the slag by-product of blast furnace ironmaking to drive a thermal decomposition reaction, making it a waste-heat-to-hydrogen technology. This is accomplished via dry granulation of molten slag that feeds a fluidized bed reactor to effect methane–slag contact. First, the proposed process and the heat and mass balances are presented. It is found that it could produce an amount of hydrogen that is equivalent to about 20% of the reductant, depending on the iron-to-slag ratio. Then, a techno-economic analysis investigates the capital and operating costs of the process, compares the hydrogen production cost to that of other processes, and examines cost sensitivity to the prices of process inputs and outputs. This analysis suggests that the process would be suitable for on-site hydrogen production and use within a plant. In addition, using the hot slag to drive the methane decomposition would reduce hydrogen production cost by 15% compared to combusting a portion of the natural gas itself. Finally, a computational fluid dynamics (CFD) modeling study of the fluidized bed reactor examines the thermal decomposition of methane and its dependence on reaction kinetics as well as reactor design and operation. The bed operated in the bubbling regime at an average temperature between 1020 and 1060 °C and resulted in as high as 82% conversion of the methane to hydrogen, with additional optimization still possible.

Graphical Abstract

Introduction

The steel industry has been considering heat recovery from hot slag to improve efficiency (e.g., as described in [1]). Slag is often disposed in pits or, in the case of blast furnace slag, it may undergo a wet granulation process to convert the material for use in cement manufacture. In either case, the high-quality energy of the slag is severely downgraded or completely lost. The recovery of high-temperature heat from blast furnace slag would require dry granulation, for which methods have been proposed and studied (e.g., [2, 3]). Typically, the intention has been to capture the slag energy as high-temperature air (at 500–600 °C) for use in the plant, and to achieve the required properties of the slag by-product for use in cement manufacture [1, 3].

Climate change concerns have spurred interest in low-carbon fuels and feedstocks, particularly for high-temperature industrial processes that are difficult to decarbonize [4, 5]. Hydrogen is carbon free, but it needs to be produced, usually from water or a hydrocarbon. A production pathway that has gained attention recently is to decompose natural gas (which consists mostly of methane) into gaseous hydrogen (or a hydrogen-rich gas) and a solid carbon (or carbon-rich) by-product. An advantage of methane decomposition is that the carbon-containing by-product is potentially easier to store or utilize and, in addition, the revenue from utilization would potentially offset the CO₂-avoidance cost. A disadvantage is that, compared to steam–methane reforming, it produces half the hydrogen per unit methane input. Nonetheless, about 60% of the chemical energy of the methane is retained after decarbonization and, unlike combustion, a carbon dioxide by-product is avoided. When carbon dioxide utilization and/or storage (CCUS) is an option, the solid carbon can feed a direct carbon fuel cell to generate power and produce a pure CO₂ stream [6, 7].

Reviews of methane thermal decomposition technologies have been carried out. Canada’s Oil Sands Innovation Alliance (COSIA) commissioned reviews of technology pathways for natural gas decarbonization with the aim to generate hydrogen for steam production, along with value-added by-products to offset the CO₂-avoidance cost [8, 9]. These studies examined a wide range of processes including the use of catalysts, membranes, and electricity. Schneider et al. [10] reviewed the state of the art in decarbonization of natural gas by thermal, electrical, and catalytic methods, including fluidized bed configurations.

The United States Department of Energy (DOE) commissioned a study and testing [11] on the production of hydrogen and carbon from thermal decomposition of natural gas and other hydrocarbon fuels. The carbon produced in the process acted as a catalyst for the decomposition reaction. Various concepts and designs for the hydrocarbon pyrolysis reactor were evaluated and tested, including packed bed, tubular, free volume, fluid wall, and fluidized bed configurations. The hydrogen concentration in the effluent gas varied in the range of 30% to 90% by volume, depending on operating parameters and the hydrocarbon feedstock. The study concluded that the fluidized bed reactor was the most suitable for decomposition of both methane and propane.

The production of hydrogen using hot slag represents a conversion of waste heat to chemical energy, as opposed to its physical capture in heated air. Li [12] showed that combined physical and chemical conversion of the slag energy is more efficient than physical capture alone, both for quantity of energy captured and for making the highest quality use of it. In addition, Barati et al. [13] showed that recovery of slag energy in chemical form offers a higher energy density than in physical form and a much better ability to transport the energy. Purwanto and Akiyama [14] performed experiments to study the kinetics of the decomposition of methane and carbon dioxide (representing biogas) using granulated slag to produce hydrogen and carbon monoxide. Their apparatus was an electrically heated, fixed bed of about 2 g of slag particles, whose average diameter was 2 mm. They observed a catalytic effect from the presence of the slag particles that was higher than for an iron catalyst but not as high as for nickel and cobalt catalysts.

This work proposes using the hot slag from blast furnace ironmaking to drive the endothermic methane decomposition, using a fluidized bed to effect contact between slag and natural gas. This requires that the molten slag undergo dry granulation. As mentioned, the scientific literature reports both experimental and modeling studies of dry granulation by rotary atomization [1,2,3]. This type of granulation method is envisioned for the process proposed here, although the granulation itself is not examined in detail. This work presents a modeling study of the heat and mass balances of the system, and a CFD analysis of the fluidized bed reactor. Despite beyond the scope of this work, small-scale or pilot-scale test results would play a critical role in refining the model and assessing the efficacy of this process at the current and higher scales.

Proposed Process

This process granulates liquid slag from blast furnaces, transforming it into solid particles. The hot particles then interact with upward-flowing methane in a fluidized bed, driving thermal decomposition to produce gaseous hydrogen and solid carbon.

Process

Figure 1 shows the proposed process. The rotary atomizer granulates liquid slag (which typically exits the blast furnace at approximately 1500 °C) to solid particles at 1300 °C, with expected diameters ranging from 100 µm to 3 mm [1]. The granulated slag falls into a fluidized bed and contacts methane traveling upward from the bottom. The methane undergoes thermal decomposition to gaseous hydrogen and solid carbon. These products flow through a heat exchanger where they cool to 500 °C, quenching the decomposition reaction and heating the incoming methane. Most of the slag particles, along with deposited carbon, are discharged at the bottom of the fluidized bed. Some of the slag particles are entrained by the gas flow, exit the top of the fluidized bed, and are separated from the product stream in cyclones. The product gas is cooled to 150 °C in a gas cooler and then entrained solid carbon is separated from the gas stream using a fabric filter. Hydrogen is separated from any remaining methane using a palladium membrane. The remaining methane is returned to the fluidized bed, where it is combined with the supplied methane prior to entering the heat exchanger. In Fig. 1, the rotary atomizer and the fluidized bed reactor are schematically separated. However, these devices would need to be integrated to facilitate the falling of high-temperature slag granules into the bed. This integration will be illustrated in “CFD Analysis of the Reactor Concept” section.

Fig. 1

Proposed process to produce hydrogen from methane using molten slag, as represented by the process simulation

In many slag-producing processes (whether blast furnace, basic oxygen furnace, or electric arc furnace), slag discharge is not continuous. The slag transfer to the proposed process would need to have buffer capacity for it to operate continuously. It is also acknowledged that fine slag or carbon particles could lead to fouling of the heat transfer surfaces in the heat exchanger. This potential fouling could be addressed by a combination of proper arrangement of the heat transfer surfaces, adjustment of flow speed, and “soot blowing.” These are beyond the scope of the present study.

Analysis and Discussion

Aspen Plus® was used to analyze the process shown in Fig. 1 for a case corresponding to the blast furnace production (particularly, the slag production rate and temperature) of a Canadian steel plant. The results are shown in Table 1. The decomposition of methane is endothermic, as indicated by its standard heat of formation, − 75 kJ/mol (the negative value implying that heat is released when methane forms from elemental carbon and hydrogen). The sensible heat transferred when lowering the temperature of 367,000 tonnes/year of slag from 1300 to 805 °C (the temperature below which methane decomposition no longer occurs and, therefore, the minimum temperature in the reactor) is 8.7 MW. (This analysis neglects the cooling of the slag and the latent heat release associated with the dry granulation; the analysis begins with solid, 1300 °C slag entering the fluidized bed.) The sensible heat from the slag performs two tasks: it drives the thermal decomposition reaction and it maintains the reactor at its operating temperature. The total energy required to convert 37,230 tonnes/year of methane is 5.54 MW. This generates 9354 tonnes/year of hydrogen and 27,868 tonnes/year of carbon black, with 8 tonnes/year of methane remaining unconverted because of the limited efficiency of the membrane gas separator. The remaining 3.16 MW maintains the reactor at its operating temperature. In Table 1, the 27,868 tonnes/year of carbon black have been divided into two parts: carbon that exits with the gas stream (and can, therefore, be captured) and carbon that deposits on the slag. This assumed division of carbon black has economic consequences that will be discussed in “Techno-economic Analysis” section. The methane conversion rate of the fluidized bed is assumed to be 80% in this process simulation. This conversion percentage will be discussed at the end of “Techno-economic Analysis” section and examined in “CFD Analysis of the Reactor Concept” section.

Table 1 Hydrogen production using slag from a Canadian blast furnace

It is found that the proposed process could produce an amount of hydrogen that is equivalent to about 20% of the reductant, assuming a typical slag-to-iron ratio of 0.3 for a blast furnace [15]. If, for every tonne of iron produced, approximately 0.3 tonnes of slag are generated at temperatures ranging from 1450 to 1650 °C then, conservatively, this slag could produce about 11 kg of hydrogen, which constitutes approximately 20% of the 50 kg needed to produce 1 tonne of iron. This percentage is provided as an approximate measure of the amount of hydrogen that could be produced. It does not necessarily imply that this quantity of hydrogen could be used directly in the unit operation that was used to generate it. For example, the replacement of carbon by hydrogen in a blast furnace depends on factors including the role of coke in supporting the burden, the impact of hydrogen on blast furnace temperature, the reduction reaction rates using hydrogen, and top gas dew point. There are other alternative uses of hydrogen within a steel plant to reduce carbon dioxide emissions.

Techno-economic Analysis

This analysis builds upon insights from existing literature and modifies their analysis for the current context. Costs adopted from literature sources are also adjusted to align with the scale and specifics of the present system.

Method

Keipi et al. [16] studied the cost of hydrogen production by thermal decomposition of methane (TDM) using a fluidized bed reactor, which they called a “regenerative heat exchanger reactor” (RHER). The main difference between the present process and that of Keipi et al. [16] is that the thermal energy for this decomposition process comes from the hot slag rather than from burning a portion of the natural gas. Like the present system, their process filtered carbon from the gas stream and separated hydrogen from methane using a palladium membrane filter. Results of their capital cost analysis are shown in Table 2, along with the corresponding equipment of the present system.

Table 2 Estimated cost of Keipi et al. [16] for thermal decomposition of methane (TDM) with hydrogen production of 529 tonnes/year (60 kg/h), along with the corresponding equipment of the present system

Keipi et al. [16] obtained their costs from the published literature for similar equipment. For their regenerative heat exchanger reactor (in Table 2, the 2.74 million EUR for capacity of 60 kg of hydrogen/h), which is similar to our fluidized bed reactor and is the dominant capital cost of their system, their cost refers to similar equipment in the power industry. They state that “costs for the process equipment were based on nominal price approximations in existing similar devices found in the power industry, such as…thermal conversion reactors, circulating fluidized bed reactors…” [17] and that “cost distribution for total capital expenditure (CAPEX) was assumed to be similar to a typical power plant [engineering, procurement, and construction], which includes equipment costs, process costs, automation and electrification costs, civil costs, and project costs (e.g., project management, engineering, and start-up)” [17].

The costs from Keipi et al. [16] were scaled according to equipment capacity using the relationship recommended by Towler and Sinnott [18]:

$$\text{Cost A} =\text{Cost B} \times {\left(\text{Capacity A}/\text{Capacity B} \right)}^{n},$$

(1)

where $n$ is typically between 0.4 and 0.9, implying economy of scale [18]. This relationship is expected to provide capital cost estimates within ± 50% [18]. Table 3 provides the scaled capital cost for the present system based on this scaling relationship, converted to Canadian dollars. The values of the scale exponent, $n$, are as recommended by [16, 18, 19] for the corresponding equipment identified in Table 2.

Table 3 Estimated capital cost for major equipment of the present system with hydrogen production of 9354 tonnes/year (1067.8 kg/h)

To convert the total capital cost to an annual cost, the standard capital recovery factor (CRF) was employed:

$$\text{CRF}=i{\left(1+i\right)}^{N}/\left({\left(1+i\right)}^{N}-1\right),$$

(2)

where $i$ is the annual interest rate and $N$ is the investment period in years.

Keipi et al. [16] recommended an annual operation and maintenance (O&M) cost of 2% of the annual capital cost, except for the membrane reactor for which they recommended an annual O&M cost of 4% of its total capital cost. Their O&M cost did not include the costs of electricity or methane, which were accounted for separately. This approach is employed in the present analysis.

Additional assumptions were required for the cost estimates. Again, natural gas was represented by 100% methane. The investment period was assumed to be 15 years, with no remaining value at the end. The annual interest rate was 10%. Given that the scaling relationship of Eq. (1) is expected to provide a capital cost estimate within ± 50% and that the proposed system is novel, and to address the possible need for other minor equipment, a 50% contingency was added to the capital cost estimates, which is higher than the 30% used by Keipi et al. [16]. Based on information from Canadian steel producers, the sale price for slag that is uncontaminated by carbon (and, therefore, useful for cement production) was assumed to be CAD 25/tonne and the sale price for slag that is contaminated by carbon (and therefore, only useful as aggregate) was assumed to be CAD 5/tonne. Like Keipi et al. [16], it was assumed that 80% of the carbon black would deposit on the slag, with the rest entrained by the product gas stream. This split of carbon black between product gas and slag is based on experiment using a laboratory-scale reactor [20]. The carbon could be removed from the slag; whether this was worthwhile would depend on economic considerations (the cost of removing the carbon versus the potential increased value of the slag) and environmental considerations (additional carbon dioxide emissions associated with, say, oxidizing the carbon or producing steam to react with the carbon). This is beyond the scope of the present study.

Results and Discussion

Results of the analysis are shown in Tables 4 and 5, which present the net annual hydrogen production cost—capital, maintenance, and natural gas costs minus the slag and carbon-black revenue. Table 4 examines the effect of natural gas purchase price on hydrogen production cost, and Table 5 examines the effect of carbon-black sale price on hydrogen production cost. In all cases, the capital cost with contingency was CAD 41.8 million and the capital cost recovery factor was 0.131 (corresponding to the 10% annual interest rate and 15-year investment period), for an annual capital cost of CAD 5.50 million. The annual O&M cost was CAD 83,500. The annual natural gas input was 37,230 tonnes/year, the annual output of slag plus deposited carbon black was 389,295 tonnes/year, the annual carbon-black output was 5573 tonnes/year, and the annual hydrogen output was 9354 tonnes/year.

Table 4 Hydrogen production cost estimate at different natural gas purchase prices with fixed carbon-black sale price

Table 5 Hydrogen production cost estimate at different carbon-black sale prices with fixed natural gas purchase price

As mentioned earlier, throughout Tables 4 and 5, it was assumed that 80% of the solid carbon would deposit on the slag, resulting in 6% carbon-in-slag and, consequently, the slag sale price would drop from CAD 25/tonne to CAD 5/tonne. There are interesting economic considerations with respect to the carbon and the slag. For example, with a carbon-black sale price of CAD 0.7/kg, the reduction in the sale price of the slag from CAD 25/tonne to CAD 5/tonne (367,000 tonnes × CAD 25/tonne – 389,295 tonnes × CAD 5/tonne = CAD 7.2 million lost revenue) would result in a net income reduction of CAD 3.3 million as the sale of carbon black (CAD 700/tonne × 5573 tonnes = CAD 3.9 million) failed to compensate for the reduction in slag revenue. If the hydrogen was to be sold, the mark-up over its production cost would have to be CAD 0.35/kg (CAD 350/tonne × 9354 tonnes = CAD 3.3 million) to compensate for the reduction in slag revenue. In the extreme, if neither the carbon black nor slag had any value (representing a CAD 9.2 million loss of slag revenue), the hydrogen mark-up would have to be CAD 0.98/kg to compensate.

The estimated hydrogen production cost varied from CAD 0.52/kg to CAD 2.69/kg, driven strongly by the price of natural gas, which was varied from CAD 1.55/GJ to CAD 11.38/GJ. Natural gas price constituted 40% to over 80% of the hydrogen production cost—the higher the natural gas price, the higher its percentage of the production cost. The O&M cost accounted for less than 5% whereas the capital cost accounted for 20% to 60%. The carbon-black sale price had a relatively small impact on the net hydrogen production cost.

The same cost estimate is graphically conveyed in Fig. 2 as a family of lines representing hydrogen production cost versus natural gas purchase price for a range of carbon-black sale prices. The slope of the lines represents the increase in hydrogen production cost per increase in natural gas purchase price, all other parameters remaining constant (i.e., the sensitivity to natural gas price). If one extends the line for zero carbon-black sale price to intersect the vertical axis, it suggests a 0.39-CAD cost/kg of hydrogen that does not depend on the price of natural gas.

Fig. 2

Hydrogen production cost versus natural gas purchase price for various carbon-black sale prices

It is important to compare the proposed process to alternative hydrogen production processes and to estimate the benefit from capturing the slag waste heat compared to combusting a portion of the natural gas itself. Table 6 compares the proposed methane thermal decomposition process using hot slag to the methane decomposition, steam–methane reforming, and water electrolysis scenarios presented by Keipi et al. [16]. Their water electrolysis scenario assumed 100% renewable energy and their thermal decomposition reaction was driven by combusting a portion of the natural gas itself. This comparison assumed 15 years of operation, a 10% annual interest rate, a natural gas price of CAD 0.463/kg, an electricity price of CAD 45/MWh, a CO₂ emissions cost of CAD 15/tonne of CO₂, and no sale of carbon black or slag. Labor cost was included in the operation and maintenance (O&M) cost. The O&M cost was 2% of capital cost for methane decomposition when combusting a portion of the natural gas itself, 5% for all steam–methane reforming processes, and 3% for water electrolysis. Delivery cost was not included. It was found that the proposed slag-to-hydrogen process is competitive with large-scale steam–methane reforming with integrated CCS and much less costly than small-scale steam–methane reforming. This supports a conclusion similar to that of Keipi et al. [16] that it would be most suitable for on-site hydrogen production and use at a plant. Comparing the Slag-TDM and NG-TDM costs in Table 6, the present process using slag waste heat reduced the hydrogen production cost by 37% compared to methane thermal decomposition driven by combusting a portion of the natural gas itself. However, the Keipi et al. [16] thermal decomposition case was at a substantially smaller production rate, so part of this cost reduction is due to economy of scale. Correcting for scale, i.e., after scaling-up the NG-TDM capital cost but retaining its full natural gas requirement, it is found that a 15% reduction in the net production cost per tonne of hydrogen is achievable by utilizing the hot slag. That there is a significant cost saving potential by capturing the slag waste heat rather than combusting a portion of the natural gas itself is consistent with the finding that the cost of natural gas is a significant contributor to the overall cost, even though about 80% [16] of the natural gas is the feedstock rather than the energy source. As another way to consider it, if natural gas is 70% of the cost/kg of hydrogen (consistent with our earlier finding for the 9354 tonnes/year hydrogen production rate) and 20% of the natural gas is used for heat, then saving that 20% natural gas saves 20% of 70% of the cost, which is 14%.

Table 6 Hydrogen production cost of this process (Slag-TDM) compared to that of the processes presented by Keipi et al. [16]

In the Keipi et al. [16] study, the CO₂ generated from the combustion of natural gas to generate steam for steam–methane reforming was not assumed to be captured. This represented 40% of the CO₂ generated. The other 60%, which came from the reforming process and was separated using pressure swing adsorption, was assumed to be captured. This is the reason their hydrogen production cost for SMR-LS + CCS depended on the CO₂ cost. Electricity for the capture facility was assumed to be 100% renewable.

The advantage of the internal recirculating gas loop is that it produces a highly purified hydrogen stream and thereby minimizes the required methane input to the system for a given hydrogen production rate. A key consideration with such a system is the conversion rate of the fluidized bed. An 80% methane conversion rate was assumed for the fluidized bed in this techno-economic analysis. It should be noted that a lower conversion rate would increase the required internal gas circulation of the system. At 80% conversion, the fluidized bed must process a natural gas flow that is 25% higher than the natural gas input to the system. At 50% conversion, the fluidized bed must process a natural gas flow that is 100% higher than (i.e., double) the input to the system. Thus, the required capacity (and, hence, cost) of this key unit operation escalates rapidly if the fluidized bed is inefficient. This emphasizes the importance of a high fluidized bed conversion rate. In addition to the fluidized bed, a lower conversion rate would increase the required capacity of the membrane reactor, heat exchanger, and solids separation (although the fluidized bed is the dominant expense). The alternative is to remove the internal gas recirculation loop and accept a hydrogen stream containing methane.

CFD Analysis of the Reactor Concept

The reactor is a pivotal unit operation in the process. Its percent conversion of methane to hydrogen is of key interest because, referring to Fig. 1, higher conversion percentage means that less methane needs to be recycled back to the reactor, resulting in a smaller reactor size for a given hydrogen production rate, less separation required at the membrane separator, and a lower gas flowrate in the loop.

Method

To examine the fluidized bed reactor concept, a CFD modeling study was performed. The technical aspects of interest were the gas and particle flow, heat transfer, thermal decomposition reaction, and fate of the solid carbon. A possible fluidized bed configuration is shown in Fig. 3.

Fig. 3

A possible rotary atomizer and fluidized bed reactor configuration. The reactor portion, indicated by the dashed lines, was examined by CFD modeling

A simplified fluidized bed reactor configuration was created for the CFD analysis. This configuration, of pilot-scale size, is shown in Fig. 4. A pilot-scale reactor had to be considered instead of a full-scale reactor because of the computational intensity of this dense solid-phase CFD simulation, which will be discussed. To simplify the model, solid slag particles were injected into the reactor at the indicated location. (The reason the solid slag particles were injected near the bottom of the reactor instead of from the top, as indicated in Fig. 3, was to provide the highest chance of success at this early design stage. It ensured the methane would encounter the hottest slag particles.) The slag discharge pipe was not used for reasons related to the computational intensity that will be explained. The reactor walls were assumed to be adiabatic. Methane was injected from the bottom and the reactor outlet was at the top. Process modeling using Aspen Plus® suggested the operating parameters shown in Table 7, which were implemented for the CFD simulation. The reactor diameter was chosen such that the fluidized bed would operate as a bubbling bed to provide a long residence time for gas–solid contact. The reactor height, as well as the location of hot slag injection, were adjusted based on preliminary CFD simulation results to reduce the reactor size and improve performance, resulting in the configuration shown in Fig. 4. Note that the fluidized bed operating temperature in Table 7 is possibly above the glass transition temperature of blast furnace slag, which would render it unsuitable for cement making. In addition to potential contamination by carbon, this is another reason the techno-economic analysis made the conservative assumption that the slag could be sold only as a construction aggregate.

Fig. 4

Simplified fluidized bed reactor configuration for CFD analysis. The diagram is not to scale

Table 7 Scaled-down fluidized bed reactor operating conditions

The methane decomposition reaction was modeled by a particle surface reaction,

$${\text{CH}}_{4}+{C}_{\text{core}}\to \left({C}_{\text{core}}+\text{C}\right)+2 {\text{H}}_{2},$$

(3)

where ${C}_{\text{core}}$ represents nucleation particles of carbon-black size but with very small mass, which were injected into the bed with the methane. Carbon was then added to these particles as the methane decomposed and, as the carbon deposited, the particle diameters were assumed to remain constant while the particle densities increased. The resulting density of carbon particles in the simulation was somewhat high but within the right order of magnitude. It will be shown that this over-prediction of carbon particle density did not affect the ability of the gas flow to suspend the carbon particles. The injection of carbon-black nucleation particles was employed to simulate the methane decomposition surface reaction and growth of carbon particles. This process must initiate at a nucleation site, and the chosen method for this simulation was to define a miniscule carbon particle. This method primarily serves as a modeling strategy, as the tracking of the surface reaction must be explicitly defined, although it could represent an actual physical requirement given that many similar processes such as condensation require nucleation sites. Note that the slag particles could also have been used to host methane decomposition and carbon growth.

No kinetic data were available for methane decomposition driven by hot slag particles in a fluidized bed, so the results presented here used kinetic data from two alternative sources. The first was based on test results from a fluidized bed using different carbon particles as bed materials [11] and the second was derived from experiments [21] in which methane was exposed to a slag surface in a graphite crucible.

Consider the experimental study of [21]. The kinetic rate coefficient is defined on a per-unit-area basis and it encodes the temperature-dependence of the reaction. The rate coefficient multiplies the surface area and methane concentration to produce the reaction rate. Two experiments were performed to measure the rate coefficient, one in which the graphite crucible was empty and one in which it contained slag (having methane-exposed areas of both graphite and slag). Comparing the two experiments, in which the surface areas of slag and graphite were well defined, enabled the determination of the rate associated with the slag surface. The temperature range of the experiments was 900 °C to 1500 °C, which contains the operating range of the present fluidized bed. Interestingly, according to this experimental study, the rate coefficient for this particular slag surface was actually lower than that for the graphite surface below 1178 °C. In any case, there is considerable uncertainty (and, likely, variability) in reaction kinetics, so it is prudent to consider both the experimental studies of [11, 21].

The multiphase flow was modeled using the dense discrete phase model in ANSYS-Fluent®, the key conservation equations accounting for continuity, momentum (Navier–Stokes), energy, and chemical species. In addition, the model included an endothermic methane decomposition reaction, using the particle surface reaction framework described in the ANSYS User Guide 2022 R2. The P1 model was used to model the radiation heat transfer and the weighted-sum-of-gray-gases model was used for the absorption coefficient of the gas phase. Particle–particle interactions were modeled using the kinetic theory of granular flow. The “Phase Coupled SIMPLE” algorithm was employed for pressure–velocity coupling, gradients and derivatives were represented using the “Green-Gauss Node-Based” scheme, pressure interpolation was set to the “PRESTO!” scheme, spatial discretization of the equations was set to “First-Order Upwind”, and the transient time-discretization was set to “First-Order Implicit.” A three-dimensional computational mesh consisting of 154,720 hexahedral cells was generated to represent the reactor. The dense discrete phase model introduced constraints on the mesh. While the accuracy of gas flow and chemical species calculations improves as the mesh is refined, the minimum cell volume must be larger than the volume of the largest parcel of particles, here having a parcel diameter of 2.4 mm. (A parcel is a particle cluster that takes advantage of particle collective motion to reduce computation.) This mesh represented a balance between the continuum and discrete phase requirements, based on testing with different meshes and parcel specifications. That the discrete phase affected the continuum phase calculation was expected, given the dominant impact of solids on the flow in this reactor.

Three discrete phases were specified for this simulation: slag particles initially in the bed, injected slag particles, and carbon particles. Preliminary simulations indicated that particles with diameter less than 140 μm would likely be blown out of the bed. Accordingly, the slag bed and injected slag consisted of particles with diameters of 150, 210, 270, 330, and 390 μm, evenly distributed. Carbon particle diameters ranged from 0.2 to 3 μm with mean diameter of 1 μm and spread factor of 3.1 according to carbon-black samples measured by CanmetENERGY-Ottawa. The particles participated in heat transfer by convection and radiation.

Table 8 provides key heat transfer parameters for the particles. The slag particle heat capacity was calculated using Aspen Plus® from the blast furnace slag composition at 770 °C. The chemical composition was 37.37% CaO, 36.15% SiO₂, 11.59% MgO, 10.09% Al₂O₃, 1.39% S, 1.05% TiO₂, 0.73% FeO, 0.48% K₂O, 0.26% Na₂O, and 0.24% MnO. No specific heat capacity data for carbon black at high temperature were found, so the specific heat of graphite at 1000 K was used [22]. The endothermic heat of methane decomposition refers to the standard heat of formation of methane. This value is used for the usual thermochemical analysis that accounts for standard heat of formation and the sensible heat with respect to standard condition. The fraction of thermal decomposition heat taken from the carbon particle and the radiative surface emissivity of slag and carbon particles were based on previous experience modeling reacting particles.

Table 8 Particle heat transfer parameters

To initialize the simulation, the reactor contained 7.9 kg of slag particles at a temperature of 1083 °C. The initial gas velocity was 0.22 m/s and the initial gas temperature was 500 °C. The CFD model simulated 16 s of operation time and achieved steady state with respect to fluidization and methane decomposition. These 16 s of simulated operation took 26 days to compute using an HP Z840 workstation using 24 cores. The relatively short, simulated operation time, due to the computational intensity, meant that the simulation could not achieve steady state with respect to solids inventory in the bed. It was, therefore, decided to not allow solids to exit the reactor through the discharge pipe. As mentioned, hot slag particles entered through the injection location to compensate for heat consumed by methane decomposition. As a result, the bed solids inventory increased by 2% during the 16 s. In real operation, the slag discharge rate would be controlled to match the injection rate. This could be accomplished in a CFD simulation, albeit at high computational cost.

Results and Discussion

The CFD simulation results confirmed that the fluidized bed operates in the bubbling regime, which is consistent with the earlier process design. The bubbling bed was found to have a height of approximately 1 m. A plot of instantaneous particle volume fraction is shown in Fig. 5. The average bed temperature at steady state remained between 1020 and 1060 °C, which is close to that predicted by the Aspen Plus® simulation (referring to Table 7), and the average temperature of the gases exiting the reactor was approximately 850 °C. In the configuration studied here, the resulting mole fraction of hydrogen depended noticeably on the decomposition reaction kinetics employed. Using the kinetics from [11], the hydrogen mole fraction was 0.7, whereas using the kinetics from Kashiwaya and Watanabe [21], it was 0.9. The 0.7 hydrogen mole fraction implies a 53.8% conversion of methane to hydrogen (each mole of converted methane results in 2 mol of hydrogen) whereas the 0.9 hydrogen mole fraction implies an 82% methane conversion. As discussed at the beginning of this section, higher conversion percentage means that less methane needs to be recycled back to the reactor. The more methane that needs to be recycled back, the larger the required reactor for a given hydrogen production rate. There is clearly a significant difference between the conversion percentages involving the two sets of reaction kinetics. A significant difference among reaction kinetics from the literature is not completely unexpected, given that the decomposition of methane is influenced by catalytic effects that depend on the composition of the material in contact with it, particularly when the material contains metal compounds.

Fig. 5

Instantaneous solid volume fraction of the fluidized bed at 16 s of simulated operation time. The slag injection location is circled. The height (m) is indicated

Figure 6a shows the predicted hydrogen mole fraction (the balance being methane) in the reactor when kinetics from [11] were used. The hydrogen mole fraction reached approximately 0.7 at the reactor midpoint and then the reaction stalled. Figure 6b shows the predicted gas temperature. As expected, the heating by slag particles occurred mostly in the bottom 1-m of the reactor, where the bubbling bed is located. The decomposition reaction continued to progress above the bed for another 2 m until it stalled. The temperature plot indicates that, as the endothermic decomposition reaction proceeded, the gas temperature dropped to a level at which the reaction rate became very slow. Thus, little was accomplished in the top 3 m of the reactor. This was confirmed by examining plots of the reaction rate. It is possible that, with further optimization of the design and operating conditions, the methane decomposition percentage could be improved. Increased bed height would create a longer high-temperature region, which would likely improve the methane conversion.

Fig. 6

Instantaneous a hydrogen mole fraction and b gas temperature (°C) at 16 s of simulated operation time (note the 7-m reactor height has been compressed by half in this figure)

Figure 7 provides the methane conversion as a function of gas residence time where, again, kinetics from [11] were used. The average upward gas velocity in the reactor is approximately 1.42 m/s, which links residence time to height. The methane conversion rate is rapid near the bottom of the reactor. At approximately 0.25 m from the bottom, when the residence time is approximately 0.17 s, the conversion of methane is 30%. At 1.42 m from the bottom, when the residence time is 1 s, 44% of the methane has been converted. The methane conversion continues until approximately 2.8 m from the bottom, corresponding to a residence time of 2 s, when 53.8% has been converted. Beyond this location, where the gas temperature has fallen to 970 °C, there is no increase. This indicates that the methane conversion rate is highly sensitive to temperature, implying that a longer residence time in this reactor would not enhance conversion. To further enhance conversion, it would be necessary to increase the operating temperature of the reactor or increase the height of the high-temperature region (i.e., increase the height of the fluidized bed).

Fig. 7

Instantaneous methane conversion versus height and gas residence time

The CFD simulations of the pilot-scale fluidized bed reactor predicted methane conversion rates of 53.8% and 82% using two different sets of reaction kinetics data from the literature. These kinetics expressions were derived from laboratory-scale experiments reported in [11, 21]. Reference [21] provided the kinetics expressions explicitly (although some unit conversion was required for our purposes) and Reference [11] provided conversion data, from which our kinetics expressions were derived. The experiments from [11] were conducted in a fluidized bed reactor using carbon particles as the bed material. The conversion rate achieved was 30.9% at a bed temperature of 950 °C and methane residence time of approximately 1 s. In comparison, our bed temperature was above 1050 °C and the methane residence time was 4.5 s. These more favorable operating conditions resulted in a conversion of 53.8%. The predicted conversion rate of 53.8% at the outlet and 44% at 1.42 m (corresponding to about 1-s methane residence time) is comparable to the experimental result of [11]. Therefore, our CFD results seem to align with the experiments reported in the literature. The experiments from [21] were conducted using a graphite crucible as well as the same crucible containing a slag surface. Although methane conversion rates were high in this experiment, and this experiment was well suited to measuring the kinetic rate, it is difficult to make a direct comparison with the present fluidized bed, which had an 82% conversion in our CFD simulation.

Although the high-temperature decomposition of methane is well-documented, there is a lack of published kinetic data specifically for the reaction in a fluidized bed reactor utilizing slag as the bed material. Existing kinetic data, such as that from [11], pertains to a fluidized bed reactor but employs carbon as the bed material instead of slag. The kinetic data from [21] are derived from experiments involving a graphite crucible containing a slag surface. These variations in bed material and experimental conditions suggest that neither set of kinetics data accurately represents the reaction in a fluidized bed with hot slag as the bed material. Conducting dedicated tests would be necessary to obtain precise kinetic data; however, such investigations are beyond the scope of this study. Therefore, this study aimed to utilize the best available kinetic data to estimate the potential operation of the proposed reactor utilizing slag as the bed material. The fluidized bed performance would almost certainly need to be improved if the conversion rate was 53.8% and a high-purity hydrogen stream was desired (requiring an internal recirculating gas loop), referring to the techno-economic discussion on the effect of conversion inefficiency on cost at the end of “Techno-economic Analysis” section.

Regarding the effect of scale, it is worth noting that methane conversion is highly sensitive to temperature. Provided the temperature remains high and uniform within the reactor, as the scale increases, high conversion rates should be achievable.

The rate of carbon particles exiting the reactor at the gas outlet was found to be 0.61 kg/h, representing essentially all the converted carbon (the mass flow rate of carbon in the methane times the 53.8% conversion rate). This implies that the flow was able to suspend the carbon particles and carry them to the gas outlet, despite their excessive density mentioned earlier. Whether the carbon particles would exit through the gas outlet in reality, given that carbon might stick to the slag particles instead and exit with the slag discharge, requires further (and likely experimental) investigation.

Conclusion

A new process was presented for thermal decomposition of methane to gaseous hydrogen and solid carbon in which a hot, dry-granulated slag feeds a fluidized bed reactor to effect methane-slag contact and drive the methane decomposition. It is found that it could produce an amount of hydrogen that is equivalent to about 20% of the reductant, depending on the iron-to-slag ratio of the process. A techno-economic analysis was performed, suggesting that the proposed hydrogen production process would be competitive with large-scale steam–methane reforming with integrated CCS and much less costly than small-scale steam–methane reforming, making it suitable for on-site hydrogen production and use within a plant. In addition, using the hot slag to drive the methane decomposition would reduce hydrogen production cost by 15% compared to combusting a portion of the natural gas itself. A CFD analysis was carried out to examine the fluidized bed reactor. It suggested that the fluidized bed approach is an appropriate method to provide methane–slag contact. It operated in the bubbling bed regime at average bed temperature between 1020 and 1060 °C, and resulted in 53.8% to 82% conversion of the methane to hydrogen. At the lower end of this conversion range, a high-purity hydrogen stream might not be economical and a hydrogen stream containing methane would have to be utilized. Further work would be required to optimize the design and operation, and to control the solids inventory in the fluidized bed. Further experimental investigations would be required to reduce the uncertainty of the chemical kinetic rate for methane decomposition, to determine the fate of the produced carbon, and to address the challenge of dry granulation.