Toward Rational Design of Nickel Catalysts for Thermocatalytic Decomposition of Methane for Carbon Dioxide-Free Hydrogen and Value-Added Carbon Co-Product: A Review

Graphical Abstract

This review evaluates data on catalyst development for thermocatalytic decomposition of methane. We present correlations between catalyst characteristics, catalytic activity (TOF), and catalyst stability (TON), notably for mono-metallic nickel and iron catalysts, which show a correlation between metal particle size and catalyst activity and stability.

Abstract

Thermocatalytic decomposition of methane provides opportunities for hydrogen (H₂) production with no emission of carbon dioxide. However, high-value carbon products need to be produced for economic deployment of thermocatalytic decomposition and to achieve a minimum H₂ selling price below the U.S, Department of Energy target of $ 1/kg H₂. In this review, we re-evaluate data on catalyst development reported in the literature and propose correlations between catalyst characteristics, catalytic stability, and properties of carbon co-products. In the first part of the review, growth mechanisms for carbon nanotubes using state-of-the-art chemical vapor deposition are reviewed to catalog the effects of catalyst characteristics, the influence of carbon sources, interactions between metal particles and supports, and metal particle sizes on carbon growth. In the second part, representative developments in mono-, bi-, and tri-metallic nickel catalysts are highlighted. We present kinetic analysis of reactions catalyzed by mono-metallic nickel catalysts, which generates a correlation between metal particle size and catalyst stability. Rational design of Ni-based catalysts for TCD of methane requires attention to the size of the metal particle and effective normalization of the reaction rates. Further attention to the distribution of the metal particle sizes may help identify catalyst properties that contribute longevity and selectivity to processes that use them. While it is tempting to focus on the highest valued carbon products (e. g., CNTs and CFs), analysis of the markets for other carbon products suggests that a more flexible approach may generate comparable returns without the risk associated with specialization.

Introduction

Because of its abundance, high hydrogen/carbon ratio, and easy handling,¹ natural gas is the feedstock for about half of the overall worldwide H₂ production. The production processes include steam reforming of methane (SMR), partial oxidation of methane (POM), and perhaps one day, dry reforming of methane (DMR).² Among them, SMR (reaction 1 below) is the most widely used to generate H₂ and carbon monoxide (CO) on industrial scales. It is often integrated with the water-gas-shift reaction (WGS, equation 2) to convert CO and steam into CO₂ and additional H₂. The net reaction for SMR (equation 3) requires a significant input of energy because it is strongly endothermic. If combustion of fossil fuels provides the input energy, then generation of additional CO₂ cannot be avoided. Dry reforming of methane (equation 4) and POM (equation 5) produce CO that must be removed if the product stream is destined to contact a catalyst based on Group VIII metals to avoid poisoning and reversion of the CO into methane.

Thermocatalytic decomposition (TCD) of methane, an alternative process for converting methane into H₂, forms solid carbon as a co-product (equation 6). See the Supplemental Information for an overview of other processes for making carbon from hydrocarbons. Because TCD is modestly less endothermic (37.5 kJ/mol H₂ vs. 41 kJ/mol H₂), it requires less energy input than the net SMR reaction. Although TCD consumes twice as much methane as SMR per unit of H₂ produced, it offers a path to generating H₂ that could be practiced (using renewable energy as the source of heat) to avoid the production of any CO₂, thus converting a fossil fuel (e. g., methane) into a feedstock that could be employed without increasing emissions of greenhouse gases.

$$ (1)

$$ (2)

Combining Equations (1) and (2) is the idealized net reaction for methane steam reforming:

$$ (3)

$$ (4)

$$ (5)

$$ (6)

Several reviews have qualitatively summarized improvements in catalyst activity and stability for TCD of methane. For example, Abbas et al.³ reviewed the effects of catalyst preparation methods, operating conditions (e. g., temperature, pressure, space velocity, etc.) and reactor types on catalyst performance. Li et al.⁴ and Qian et al.⁵ investigated reaction mechanism and kinetics on diverse metal catalysts to illustrate the effects of catalyst properties on catalyst deactivation. Zhang et al.^1a and Hamdani, et al.⁶ reviewed carbonaceous materials (carbon as catalyst or as support) and their application in TCD of methane, which gives a way to avoid catalyst regeneration. Ashik et al.⁷ and Muhammad et al.⁸ detailed the effects of co-feeding hydrocarbons on improving catalyst stability in TCD process. In addition, there have been reports of progress in reactor designs. For example, McFarland and co-workers⁹ used a molten salt reactor to enhance catalyst stability. Reactors using substituted heating sources such as plasma and solar to electrical heating were also been reviewed elsewhere.^{5, 10}

However, few reports quantitatively correlate activity, longevity, and selectivity of catalysts with their measurable catalyst properties such as particle size, particle composition, or support characteristics. Here we provide such correlations for activity and longevity. Activity, when expressed as a turnover frequency (TOF) at a reference temperature exhibits a clear linear dependence on the size of the catalytic domains for supported metal catalysts. We provide an explanation for why longevity is more difficult to relate to catalyst properties. Finally, we resort to anecdotal explanations for carbon co-product selectivity (i. e., type of carbon that is produced). Devising structure-activity correlations for selectivity still appears to be out of reach.

Activity

Even though reporting catalytic rates normalized by a count of the ostensible, active site was introduced to the field of heterogeneous catalysis nearly 70 years ago,¹¹ there is still a paucity of TOF data reported in the literature of TCD of methane. We could find only a few explicit measurements of TOF measured by a surface titration¹² and several others for which TOFs could be inferred from the reported diameters, d, of the exposed metal particles combined with specific rates (i. e., per unit mass of catalyst). We combined those data to estimate fractional exposures,^12b FE=1.01 nm/d, and hence TOFs (Table 1). However, we therefore omitted in the analysis below more than 50 papers where only specific rates were reported. We appeal to authors to include and for reviewers and editors to insist on including sufficient information in future publications to permit estimating TOF information, which is an obligatory characteristic of any catalyst.¹³

Table 1. Catalysts whose TOFs for TCD of methane could be estimated.

Catalyst family	D, nm^a	Range of TOF, Hz^b	References
Ni/oxides	7–21	0–2.8	Xu et al.,^12a
Ni/oxides	12–1,629	1.7–260	Chen et al.,^12b
Ni/MgO	17–25	0.7–2.0	Wang and Baker^12c
Ni/oxides, graphite	20–28	0.07–0.15	Takenaka et al.,^14
Ni/MgO	9.6–68	0.04–0.19	Echegoyen et al.,^15
Ni/promoted SiO2	5–968	0.1–22.3	Zapata et al.,^16
Ni/Cu+ Al2O3	20–24	0.15–0.16	Bayat et al.,^17
Ni/aluminosilicates	7–250	0.1–12	Choudhary et al.,^12d, 12e
Ni/La2O3	15	1.6	Gallego et al.,^18
Ni/CeO2	22–30	2.3–3.3	Li et al.,^19
Ni/TiO2+Al2O3	11–22	1.8–3.9	Awadallah et al.^20
Fe/SiO2, Al2O3, ZSM-5	4–16	0.1–0.2	Wang et al.^21
Supported, promoted Fe	5–103	0.0004–0.12	Zhou et al.^22
Ni+Fe/SiO2	9–29	0.02–1.0	Kutteri et al.^23

• a) Mean particle size of the domain of the indicated metal. b) Extrapolated from measured rates to 1,000 K using an apparent activation energy of 65 kJ/mol.

Our own work^12a and that of Takenaka et al.¹⁴ showed deviations from the simple relationship between average particle size determined from X-ray diffraction or transmission electron microscopy (TEM) and fractional exposure, determined from a chemisorption or average coordination number of the metal atoms. For example, Takenaka et al.¹⁴ employed catalysts whose mean particle sizes, from a Scherrer analysis of the X-ray diffraction powders patterns, were 20–28 nm. However, their extended X-ray adsorption fine structure analysis of the samples gave average coordination numbers of the metal atoms in the 4.5–10.5 range. It is straightforward to show from a correlation devised by Jentys²⁴ and counting statistics tabulated by van Hardeveld and Hartog²⁵ that such small coordination numbers correspond to fractional exposures in the 100 %–40 % range instead of numbers in the 3.5 %–5 % range calculated from the mean particle sizes. It is not clear to us what the discrepancy signifies nor how it might correlate with the activity of the catalysts. We can only suggest the need for additional research.

The included initial TOF exhibits a strikingly simple dependence on particle size (Figure 1), in which the effects of promoters and supports are, to first order, negligible. However, these results are difficult to analyze because close examination reveals that the cited kinetic studies are performed under conditions that are near thermodynamic equilibrium (i. e., near equilibrium conversion). Moreover, the apparent structure sensitivity could also arise from (internally) mass transport limited kinetics, if larger particles of the catalytically active metal segregated towards the surface of the support particles.

Figure 1

Dependence of initial TOF for TCD of methane (time-on-stream=0) on the domain size of the indicated metal. TOF extrapolated to 1,000 K using an apparent activation energy of 65 kJ/mol. The literature results used in this analysis weree obtained from different works and are represented by different symbols : + Xu et al.,^12a × Chen et al.,^12b ▪ Wang and Baker,^12c ▾ Choudhary et al.,^{12d, 12e} ▴ Takenaka et al.,¹⁴ • Zapata et al.,¹⁶ ⧫ Bayat, et al.,¹⁷ * Gallego et al.,¹⁸ ◂ Li et al.,¹⁹ ○ Awadallah et al.,²⁰ ◊ Zhou et al.,²² and □ Wang et al.²¹ The experimental results are summarized in Table 1.

To test whether alloying modifies the effect of the size of the metal domains,²⁶ we employed the activity trends (the two lines) shown in Figure 1 to estimate the activities of catalysts containing both Fe and Ni, ostensibly as the completely miscible alloys.²³ The estimates were formed as the composition-weighted averages of the TOFs of the metals, TOF_Ni(d) and TOF_Fe(d), at the measured particle diameters, d, determined from a Scherrer analysis of a bimetallic catalyst with metal fraction, ϕ_i, determine by inductively coupled plasma-optical emission spectrometry:

$$

Then, the TOFs were transformed into methane conversions, X_CH4, by rearranging the equation from which a TOF is calculated from the number of exposed atoms, n_surf, and the molar inlet flowrate of the methane, Q_CH4.

$$

The parity plot contains one adjustable parameter: the estimated conversions were divided by 4 to attain unity for the slope of the plot (Figure 2). The inference is that the metals in the alloyed catalysts behave as they do in catalysts containing the pure metals at the same particle size. In other words, alloying has no special effect on initial rates of the TCD reaction.

Figure 2

Comparison of measured and predicted methane conversions (X_CH4), the latter determined from the composition-weighted, particle-size dependent TOF shown in Figure 1. Data are from Kutteri, et al.²³ Dividing the predicted X_CH4 by 4 makes the slope equal to 1.

Conveniently, the linear trend with particle size, d, implies that initial activity will not depend on the distribution of particle sizes, PSD(d), in an actual catalyst (Equation 1). When TOF(d)∝d (i. e., TOF(d)=a×d), the PSD-averaged TOF, ⟨TOF(d)⟩ is identical to the TOF evaluated at the PSD-averaged particle size, ⟨d⟩:

$$ (1)

However, as will be discussed next, the longevity of the catalysts does appear to depend on particle size and hence, likely, on the particle size distribution.

Longevity

Inspection of published trajectories of rates of TCD as a function of time-on-stream show at least three different behaviors (Figure 3): long-term stability, finite stability, and rapid decay. Of these behaviors, long-term stability is clearly the most desirable for the production of industrially relevant quantities of H₂ and carbon co-products.

Figure 3

Notional modes of decay of TCD catalysts using arbitrary units.

Finite stability can be a consequence of filling pores in the support with the produced carbon and can be addressed by employing a nonporous¹⁴ or a planar support.²⁷ Rapid decay can arise from self-poisoning (i. e., coking) or metal sintering (i. e., loss of surface active sites).

There are fewer data available than for the correlation for initial activity, so it is difficult to discern a simple trend with particle size for catalyst longevity. We equated the latter to the turnover number (TON) provided by exposed surface sites, TON_surface (Figure 4).

Figure 4

Dependence of catalyst longevity on particle size of the metal particles. The number of surface atoms used to normalize the TON that is used to calculate the TOF for each sample. The literature results used in this analysis were obtained from different works and are represented by different symbols: + Xu et al.,^12a × Chen et al.,^12b ▪ Wang and Baker,^12c ▴ Takenaka et al.,¹⁴ • Zapata et al.,¹⁶ ⧫ Bayat, et al.,¹⁷ and ◂ Li et al.,¹⁹ The experimental results are summarized in Table 1.

According to Chen et al.,^12b carbon yield goes through a maximum for Ni particles around 40 nm in diameter (i. e., rapid decay minimizes in that size range). They ascribe the behavior to the driving force for transporting carbon through the catalyst particles from where the methane dissociates to where the nanostructures form. Because that trend was not evident from Figure 4, we performed a simple, macro-kinetic analysis (Equation 2) of readily accessible rate data. The analysis, presented below, supports the idea that the rate of deactivation decreases (and hence longevity increases) as the size of the supported metal particles increases.

$$ (2)

The quantities employed are the methane in the reactor, [CH₄]; the empty surface sites on the catalyst, A; the sites that have been carbided, B; the carbon co-product that is formed, G; and the deactivated sites, X. We fitted the space velocity, Q, and the rate constants, k₁ through k₃, to rate data (Figure 5) from two sources, Xu et al.^12a and Chen et al..^12b It was necessary to include the space velocity as a fitting parameter because the actual number of catalytic sites present in the experiments is never known.

Figure 5

Deactivation kinetics of Ni-based catalysts during the TCD of methane. Data in the top and middle panels come from Xu et al.^12a Data in bottom panel come from Chen et al.^12b The curves were calculated by fitting the solutions of the differential equations in Equation 2 to the data.

The rate constant that tracks deactivation of the catalysts, k₂, shows a modest trend to smaller values with increasing size of the Ni particles (Figure 6). Chen et al.^12b did not include detailed deactivation data for catalysts containing larger particles of Ni so this analysis does not show the extremum they found in the lifetime of the catalysts.

Figure 6

Estimated rate constants for deactivation of nickel catalysts in the TCD of methane.

To estimate the connection between solubility and diffusivity of carbon in metals as a function of particle size, we first correlated the measured solubility and diffusivity of C in several group VIII metals²⁸ (Figure 7) which we could express analytically (Equation 3:

$$ (3)

Figure 7

Correlation between diffusivity and solubility of carbon in Group VIII metals.

where the diffusivity, D_c, has units of cm²/s and S_c, the carbon solubility, is expressed as mol fraction.

While there are several measurements of the solubility of carbon in the bulk of metals employed as TCD catalysts,²⁹ here we are obliged to understand the solubility of carbon in nanometer-sized domains. In lieu of unavailable experimental measurements, we found that is possible to extract from molecular dynamic estimates³⁰ an analytic expression (Equation 4) for the solubility of carbon, S_c, as a function temperature; T, particle diameter, d; and the experimentally determined free energy of formation^29d of the lowest carbide, which here is Fe₃C (Figure 8).

$$ (4)

Figure 8

Solubility of carbon in Fe as a function of particle size and temperature. Left: Calculated results³⁰ plus their extrapolation to 5 nm particles (dashed line). Right: Parity plot showing that Equation 3 reproduces the trends for Fe catalysts containing 10–50 nm particles.

The parameters in Equation 4 fitted to data in Figure 8 for carbon in small particles of Fe were a=0.0315, b=10.36, and c=0.240. The expression for ΔG was that reported by Shatynski^29d for Fe₃C in the 463–1,033 K temperature range (Equation 5:

$$ (5)

The small value of c in Equation 4 suggests that carbiding the small particles requires about 25 % less energy than carbiding the bulk metal, likely a consequence of the disorder in small particles that increases with decreasing particle size (the entropy of formation can be factored out of ΔG in Equation 4 and incorporated into the pre-exponential fraction). Of course, the parameters in Equations 4 and 5 will change with the metal, but the goodness of fit, illustrated by the parity plot in Figure 8, encourages us to generalize to rationalize the existence of a particle-size dependent maximum in catalyst longevity.

An underlying assumption is that adsorption and dehydrogenation of the methane occurs on sites distal from where carbon nanotubes (CNT) or carbon fibers (CF) form and that it is diffusion of the carbon through the bulk of the particle that feeds the growing carbon structures.³¹ Diffusive flux, n=D dC/dz, is related to difference in concentration of the surface and bulk carbon: dC/dz≈(C_surface−S_c)/d where d is a length comparable to the size of the metal particles. Therefore, the flux, n, can be estimated by substituting the solubility calculated using Equation 4 into the expression for the diffusivity calculated from Equation 3. The results for Fe as a function of particle size, d, and temperature show a narrow maximum at very small particle sizes, followed by a broad, temperature-dependent minimum in the estimated flux (Figure 9). Evidently, the details of the calculations will depend on the metal, the actual surface coverage, and the actual temperature profile in the reactor.

Figure 9

Approximation of the flux of carbon through small particles of iron estimated by combining Equations 3 and 4. The surface concentration of the carbon was fixed as indicated.

Even so, the calculations broadly suggest that very small particles, d<10 nm, may deactivate quickly because their surfaces carbide (methane keeps arriving and decomposing but the carbide it forms cannot be transported through the bulk to the growing carbon structures because the flux is small). Others³² have come to the same conclusion but for a different but related reason (insolubility of carbon in the smallest particles – according to Equation 3, the two properties are highly correlated). Larger particles exhibit larger diffusive fluxes, which is consistent with their enhanced longevity. The existence of the broad minimum in the estimates of the flux apparently contradicts the results found by Chen et al.^12b (maximum carbon productivity at d ≈40 nm). Therefore, either the calculations have displaced the position of the flux maximum and/or they indicate the need to better quantify particle size distributions in the catalysts tested. Catalysts containing large particles typically have a broad distribution of particle sizes, which, depending on the mean particle size, would tend to average different regions of the flux curves.

We note that dissolution of carbon in the metal particles has been considered essential to the detachment of the particles from the stub that grows into a CNT.³³ The detachment releases the particles from the support, precluding, ordered, epitaxial growth. Dissolved carbon can also play a role in the structuring of alloy catalysts by decreasing the enthalpy of activation for diffusion,³⁴ however, as noted above, alloying and support effects play a secondary role, compared to that of particle size.

Here, selectivity refers to the type of carbon that forms: CNTs, CFs, unordered graphite, carbidic coke, and, in the case of CNTs and CFs, their different dimensions (i. e., morphology, width, length, number of walls³⁵), defects (e. g., I_D/I_G²⁰), and mode of growth²³ (tip, base). There are indications that the filamentous carbon forms preferentially on metal particles of intermediate size: neither very small particles (<10 nm)^{32, 36} nor very large particles (>100 nm)³⁶ grow filaments.

There is disagreement about the relation between the diameter of carbon structures and the diameter of the metal particles that catalyze their growth. Earlier experiments, particularly those using electron microscopy to visualize the carbon structures and the metal particles, and modeling³³ indicated that the two dimensions are closely connected.^31c However, experiments with well-defined distributions of metal particles suggest the absence of a correlation between the two dimensions.³⁶ Changes in the size and shape of the metal particle size during the growth of the carbon structures likely bears on this issue.³⁷ Growth of different morphologies, e. g., “bamboo” structures may correlate with the presence of liquid-like, highly disordered metal particles, present at higher temperatures.³⁸

Techno-economic assessments suggest that TCD of methane can become competitive with commercial SMR processes and meet the U.S. Department of Energy target of $ 1/kg H₂ if the solid carbon co-product is sold to lower the minimum H₂ selling price.³⁹ Possible carbon products include carbon black, needle coke, graphite, CFs, and CNTs. Therefore, it would seem at first glance to be of practical interest to control the selectivity of the TCD reaction with respect to the type of carbon that forms. However, if the price elasticity that applies broadly across the chemical process industry (Figure 10) also applies to the price/demand for those specialty carbons then the integrated economic value of the enterprise will not be sensitive to the type of carbon that can be made – each can find a customer at the right price. As an example, the vertical line in Figure 10 shows that using TCD of methane to add 10 % H₂ into the flow of pipeline natural gas used in a large market (viz. California) could generate a supply of carbon that could be sold at about current price of needle coke or carbon black without swamping the market for the material.

Figure 10

Global price and demand for specialty carbon overlaid on a general correlation of product elasticity in the chemical process industry. Solid black line was calculated from a similar graph constructed by Szmant,⁴⁰ updated⁴¹ to current prices using the producer price indices for chemicals⁴² in 1989 and 2019=112/255.9 and World Bank estimates of the ratio of the gross world product in constant 2010 dollars (1989 and projection for 2019)=36.9 TUSD/88.1 TUSD.

Carbon black (<0.4–2 USD/kg),^39a is mainly used as an additive in the manufacture of tires, rubber and plastic products. Needle coke (~1.5 USD/kg) is used to produce graphite electrodes for electric arc furnaces in the steel industry. Graphite (≳10 USD/kg) is used in lithium-ion batteries because of its excellent conductivity and ability to intercalate lithium metal. CFs (25–113 USD/kg) are employed in aerospace, automotive, wind turbines, construction etc., as reinforcing agents. While CNTs can also be used as additives to polymers and plastics for automotive and construction industries, their applications in electronics and energy storage are expected to grow significantly. CFs and CNTs can command much higher prices (up to 113 USD/kg and 600 USD/kg respectively)^39a if quality requirements are met. Their current market sizes are, however, limited as indicated in Figure 10. Existing commercial processes burn H₂ to produce the carbon. TCD of methane offers an opportunity to achieve high-grade production of both carbon and H₂.

Mechanisms of CNTs Growth in CVD

Theoretical Mechanisms

In the past 20 years, CCVD has been used to synthesize CNTs with high yield, purity, and CNT alignment. However, it is still an art to choosing suitable materials (i. e., catalysts and carbon sources) and processing parameters because of the complexity in CNT growth. There are general guidelines about the role of catalysts. Metallic catalysts should be able to catalyze the dissociation of the gaseous carbon precursors. The catalysts need to enable carbon diffusion and the chemical interaction between carbon and the metal particles. They are also anticipated to be an anchor for CNT nucleation and maintain a reactive edge for nanotube to grow.^38a A balance between activation of the reactant (C−H bond breaking) and the dissolution and migration of surface atomic carbon in the metal sublayer is key to the enabling active yet stable catalysts.⁴³ A similar balancing of kinetics has recently been found to permit coke-free dry reforming of methane.²

There are two generally accepted models for CNT growth: 1) vapor-liquid-solid (VLS) originally developed by Wagner and Ellis⁴⁴ for the growth of silicon whiskers and 2) vapor-solid-solid growth. The major difference in the models lies in the physical state of the metal particle (liquid vs. solid respectively). The VLS model hypothesizes that 1) the metal particle transforms into a liquid droplet as the site for carbon decomposition and 2) carbon atoms saturate the liquid particle through bulk diffusion.^38a Once oversaturation occurs, carbon atoms will precipitate on the other side of the metal surface and then start the growth of a carbon structure. Results from Baker et al.⁴⁵ on the growth activation energies suggested that bulk diffusion was necessary and was the rate determining step for carbon growth. The measured activation energies accorded with those for bulk diffusion through metals in solid state, which contrasted with the liquid particle hypothesis in the VLS model. Most recently, TEM/X-ray photoelectron spectroscopy results provided strong experimental evidence that the metal particles (Ni and Fe catalyst particles) remained in solid state but were highly deformed during CNT growth.⁴⁶

Furthermore, two types of CNT growth have been suggested: 1) tip growth and 2) base growth.^,[47,35] The difference likely depends on how strongly the catalytic nucleus binds to the support.⁴⁸ Both tip growth^{46c, 49} and base growth^{46b, 46d, 50} have been observed by in situ high-resolution TEM.

Factors that Influence CNT Growth

Operating conditions (e. g., temperature, pressure, gas flow rate, deposition time, etc.), reactor geometry, carbon sources, catalysts and supports all affect CNT growth. To pave the way for directing CNT growth based on catalyst design, we now focus on materials, namely carbon sources, catalysts, and supports.

Although TCD of the methane reaction defines methane as the carbon source, different carbon-containing compounds have been used in CCVD. Thermodynamic stability, compound composition, and molecular structure of the carbon sources affect CNT formation. Most commonly studied systems are ethylene, acetylene, methane, CO, and ethanol.^38a Among the various compounds, only methane decomposes endothermically,⁵¹ meaning that methane decomposition needs a temperature >400 °C to proceed thermodynamically. The other carbon compounds can be thermodynamically decomposed at a temperature <400 °C. Partial pressures of the gaseous species also can affect the decomposition equilibria in which higher partial pressures of the hydrocarbon will lower the equilibrium conversion; hence, inert gases are typically used to dilute the hydrocarbon streams and achieve high conversion levels. Moreover, heteroatoms (e. g., H and O) in the carbon sources can generate byproducts during the decomposition. These byproducts especially H₂ and H₂O are known to affect the growth of CNTs, as additives.^38a Furthermore, the molecular structure of the carbon source has been shown to strongly influence CNT morphology. For instance, small hydrocarbons (e. g., methane, ethylene, acetylene, etc.) were found to generate straight hollow CNTs, while larger, cyclic hydrocarbons (e. g., benzene, xylene, cyclohexene, etc.) produced curved CNTs.⁵² Therefore, it is critical to choose the carbon source and subsequent reaction conditions to form desired CNTs. In TCD of methane, co-feeding of other hydrocarbons is an effective method for improving catalyst performance. That method is discussed elsewhere in the literature.⁸

Perspective on Future Research Directions

While there already is an extensive body of work exploring the TCD of methane, most of it lacks the proper analysis needed to determine catalytic activity- structure relationships (e. g., TOF, k_deactivation) and move the field forward. To that end, we recommend future researchers focus on (1) testing the catalysts away from mass transfer and thermodynamic limitations, (2) reporting the reaction rates (instead of conversions) under controlled reaction conditions, (3) fully characterizing the catalyst before and after reaction to elucidate the deactivation mechanisms (if any) and (4) correlating the catalytic performance to a relevant property of the catalysts. As described in this work, only a few selected works reported the necessary information to do a proper kinetic analysis (Table 1); hence, we recommend future researchers to take into consideration the following points:

• Catalyst loading. The TCD reaction is thermodynamically limited at temperatures <800 °C and pressures >1 bar; hence, researchers need to pay attention not to confuse stable performance with thermodynamic (or mass transfer) limitations. This is usually the case when overloading catalyst in the reactor and not testing at high enough feed flow rates [i. e., weight hour space velocities (WHSV)] or partial pressures of methane. Researchers are recommended to always perform a few tests under different WHSV at a given reaction temperature and partial pressure of methane to assess whether the catalyst is operating under kinetically limited reaction conditions. That is, if the conversion stays similar when changing the WHSV, then the catalyst is operating under thermodynamic limitations and no kinetic analysis can be done with the results. On the other hand, if the conversion decreases proportionally to the increase in WHSV, then the catalyst is operating under a kinetic regime (at least with respect to external mass transport) and the results can be used to perform the proposed kinetic analysis.

• Normalization of rates. The reaction rates need to be properly normalized by the amount of active phase used; that is, the specific reaction rate. In the few cases that the reaction rates are normalized, the total amount of catalysts is used to estimate the specific reaction rate. However, this term is misleading when comparing to other literature values that use completely different metal weight loadings (traditionally range between the 1 to 60 wt%). A catalyst with 1 wt% of active metal has 60 times lower mass than a catalyst with 60 wt%; hence, the rate normalization must be done in terms of mass of active phase (i. e., metal).

• Reaction time. Before claiming that a catalyst is stable, enough reaction time is needed for the active sites to turn over the catalytic cycle several times. This is normally not the case when large catalyst beds are employed which coke and plug quickly due to the deposition of carbon. We recommend the researchers design the reactors or experiment in a way that catalyst deactivation can be observed before the catalyst bed plugs. Hence, the researchers are advised to use small amounts of catalysts to test for catalyst stability as well as to operate the reactor in fluidized-bed operation instead of packed bed, so the reactor is not plugged as easily.

• Metal particle size. As thoroughly discussed in our TCD work,^{12a, 38b} the metal particle is the main parameter controlling the catalytic activity and deactivation rate at a given set of reaction conditions. Hence, future researchers must pay extra attention to characterizing the metal particle size (with TEM and X-ray diffraction) before and after reaction to correlate activity to metal particle size as well as to determine the mechanism of deactivation (e. g., sintering or fragmentation of metal particles, changes in alloy composition). Furthermore, the metal particle size can be used to directly estimate the fraction of active site available for reaction (i. e., fractional exposure) and calculate the much-neededv TOF.

• Composition of active sites. Future researchers are also advised to track the composition of the catalyst and active sites during reaction because certain metals might be selectively removed from the catalyst (due to solubility) during the carbon co-product growth, resulting in dealloying of the active site and transformation of the catalysts. Besides tracking the catalyst composition, future researchers should pay extra attention to the presence of metals on the carbon co-product as well as the location of it (i. e., walls, defects) because it might provide hints to deactivation mechanism.

Furthermore, future researchers must take into consideration the potential markets for their carbon co-products derived via the TCD process. As such, we recommend comparing key properties (e. g., oxidation temperature, thermal and electrical conductivity, Raman-derived I_D/I_G ratio) against comparable existing carbon products generated via different methods to assess the market impact as well as sale price and potential revenue. Additionally, future researchers must investigate new applications for the TCD-derived carbon co-products to ensure there is a carbon market large enough to utilize all the generated carbon and avoid market saturation.

Acknowledgments

This work was financially supported by the U.S. Department of Energy (DOE), Office of Energy Efficiency and Renewable Energy, Hydrogen and Fuel Cells Program, and was performed at Pacific Northwest National Laboratory (PNNL) and in collaboration with the H2@Scale Consortium. Financial support also was provided by Southern California Gas Company and C4-MCP LLC through a Cooperative Research and Development Agreement. PNNL is a DOE national laboratory located in Richland, Washington. Some catalyst characterization studies were performed using the Environmental Molecular Sciences Laboratory, a DOE Office of Science User Facility sponsored by the Office of Biological and Environmental Research. The views and opinions of the authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed or represents that its use would not infringe privately owned rights.

Conflict of interests

The authors declare no conflict of interest.

Biographical Information

Robert A. Dagle is a Chief Research Engineer at PNNL and currently manages >$ 3 M annually in DOE-sponsored research developing new catalytic processes for the upgrading of biomass- and fossil-derived intermediates into low carbon fuels and chemicals. Current projects are aimed at producing renewable jet blendstocks, higher olefins, methanol, and butadiene, and CO₂-free hydrogen and value-added carbon nanomaterials from methane. Robert works with several academic partners that include Washington State University, West Virginia University, and Oregon State University. Robert also collaborates with commercial partners that include LanzaTech, Bridgestone, and Southern California Gas Company.

Biographical Information

Dr. Lopez-Ruiz joined PNNL in 2015 as a Senior Research Engineer at PNNL. His research interests include thermal and electro catalytic transformation of biomass- and waste-derived molecules into fuels and chemicals. He develops novel catalyst designs by combining kinetic, characterization, and computational studies. He has a double major in Chemical Engineering and Industrial Engineering from the Universitat Rovira i Virgili in Tarragona, Spain, obtained Masters in Chemical Engineering from Bucknell University in Lewisburg, PA, and received his Ph.D. in Chemical Engineering from the University of Virginia in Charlottesville, VA, where he worked with Robert J. Davis.

Biographical Information

Dr. Changle Jiang is a Research Assistant Professor in Department of Chemical and Biomedical Engineering, West Virginia University. Dr. Jiang's research focuses on clean hydrogen and carbon nanotube production via catalytic methane decomposition by microwave irradiation. He also works in catalysis and reaction engineering related to biomass conversion to platform chemicals as well as plastic upcycling to fuel. Dr. Jiang earned his Ph.D. degree in Forest Resources Science from West Virginia University in December 2019. His dissertation focused on biomass carbonization using carbon dioxide.

Biographical Information

Dr. Robert S. Weber is Senior Engineer emeritus at Pacific Northwest National Laboratory where he worked for one decade. His research activities continue to focus on heterogeneous catalysis for fuels and chemicals. Before that, he was the director of the chemical engineering practice of TIAX and a member of the chemical engineering faculties of the University of Delaware and of Yale University. Currently he serves on the Editorial Advisory Boards of ACS Omega and of AIChE's Journal of Advanced Manufacturing and Processing. He holds a BA from Cornell University and a PhD from Stanford University.

Biographical Information

Dr. Jianli (John) Hu is a professor in chemical engineering and the Director of Shale Gas Center at West Virginia University. He leads an interdisciplinary team carrying out cutting edge research in natural gas conversion and renewable energy utilization. His research interests span across the fields of reaction engineering, surface chemistry, plasma and microwave-enhanced catalytic reactions. Before joining WVU, Dr. Hu led innovation efforts at Koch Industries, Pacific Northwest National Laboratory and BP Oil. He has been granted 45 U.S. patents and published over 125 peer-reviewed journal articles, and edited 2 books.

Biographical Information

Dr. Mengze Xu obtained her Ph.D. degree in Applied Chemistry from Colorado School of Mines & the National Renewable Energy Laboratory in 2017. Her research was focused on heterogeneous catalysis for biomass upgrading. She joined Pacific Northwest National Laboratory as a postdoc researcher and mainly developed new catalyst designs to produce CO₂-free hydrogen and value-added carbon materials from methane. Currently Dr. Xu is a Senior R&D Engineer at the China National Offshore Oil Corporation (CNOOC), where she works on novel technology transformation from bench scale to pilot scale in the fields of C1 chemistry and renewable energy utilization.