Advanced manufacturing in heterogeneous catalysis

Heterogeneous catalysis is a critical backbone of the chemical industry with almost 95% (by volume) of manufacturing processes relying on the use of some sort of catalysts.¹  As of 2020, the catalyst market was valued at 35.5 billion USD and is projected to reach 57.5 billion USD by 2030 with an average annual growth rate of 4.9%.²  Tremendous progress has been made over past decades in terms of technological breakthroughs and innovations including scale-up of high-throughput experimentation, development of novel catalytic materials and processes, and increase in computational power for theoretical calculations,³  thus improving the fundamental understanding and commercialization of catalytic processes. As concerns over the environmental impacts of chemical processes have risen in importance, the focus has shifted towards improved efficiency and sustainability, and new, flexible chemical processes with milder conditions and smaller footprints (often known as process intensification). To achieve this, several fundamental concerns need to be addressed to improve the turnover efficiency of catalytic systems such as improved heat transfer, increased gas–solid contact, and lower OPEX (operational expenditure) costs.

Additive manufacturing (AM) is an important tool that has been around for more than 30 years and has acquired significant interest from both academia as well as industry. The driving factors for AM are the increased demand for customized products, shorter product development cycles, reduced manufacturing costs and lead times, and an elevated focus on sustainability.⁴  Given the advantages of customized and precisely shaped product manufacturing, this technology holds promise in the development of catalytic materials and reactor systems. Due to the very high interest and a rapidly developing field, numerous reviews^3,5–10  have been published recently and have highlighted the progress, challenges, and various applications of AM in catalysis.

American Society for Testing and Materials (ASTM) classifies AM methods into seven different categories based on the type of building and bonding of the layers of the material being fabricated. These include: Vat Photopolymerization (VP), Powder Bed Fusion (PBF), Material Jetting (MJ), Binder Jetting (BJ), Material Extrusion (ME), Sheet Lamination (SL), and Directed Energy Deposition (DED).⁸  Out of these, ME, VP, BJ, and PBF category methods will have the most profound impact on catalytic technologies^5,10  as ME offers the simple operation and low cost; VP offers high precision and rapid prototyping; BJ offers flexibility and high speeds; and PBF offers high mechanical strength and stability for metals. These methods can be further divided into Fused Deposition Modeling (FDM), Direct Ink Writing (DIW) or robocasting, Stereolithography (SLA), Selective Laser Sintering (SLS), Selective Laser Melting (SLM), and Binder Jetting (BJ). An excellent summary on the various technologies, classifications, and commercial developer brands can be found in a review by Laguna et al.⁷ 

The advancement of any technology, including catalytic, will require validation through several stages of development at different scales for its validation i.e., concept → bench-top → pilot plant, before it can be adopted for commercial manufacturing. When the technology progresses through the concept and bench-top validation, engineering considerations are needed to design the catalyst system in a commercially representative form for the appropriate reactor system and designing and developing associated downstream processes come into picture to obtain the final product. AM enables the ability to produce customized catalytic and reactor components for chemical processes across these different technology scales.

Furthermore, chemical manufacturing currently occurs in large, centralized facilities. The push for decarbonization of energy production, chemical industries, and manufacturing has presented several opportunities for chemical energy storage, flare gas conversion, or point source production where manufacturing of chemicals could occur in small, distributed scales. Systems for these applications would be markedly different than their larger counterparts and require compact, flexible processes with unique process-intensification designs to maximize heat and mass transport for more efficient catalyst performance. AM presents the opportunity to develop new and customized catalyst designs and reactor components that would lower energy intensities of these smaller scale systems.

The current chapter will cover the recent progress on the advanced manufacturing methods in heterogeneous catalysis. This includes advances in catalyst coating methods, AM of catalytic process components (catalysts/supports and reactors), and in process intensification approaches in catalytic processes. The chapter will also cover aspects of artificial intelligence and machine learning (AI/ML) in catalysis and how it helps to formulate and screen new catalysts. The impacts and challenges of all these techniques will also be covered.

The production of major chemicals such as ammonia (from hydrogen and nitrogen), methanol (from carbon monoxide hydrogenation), synthesis gases (from methane steam reforming), epoxides (alkene oxidation), fuels (from fluid catalytic cracking), light olefins, propylenes (from methanol), different aromatics and their derivatives, all require various catalysts. Catalysts are not only used in developing fuel cells and photovoltaic cells to replace conventional fossil fuels dependence, but also are indispensable for automotive exhaust, industrial effluent, and municipal waste treatments to reduce pollution and other adverse environmental impacts. However, the demands of reaction conditions for chemical synthesis are rigorous requiring catalytic materials to be designed to produce high conversions and long-term selectivities for their respective process under conditions of high temperature and pressure with oxidizing and/or reducing feedstocks. Single component catalysts are often unable to produce reliable activity for any reaction, and in turn, require the use of multicomponent materials to improve stability, activity, and robustness. There are different types of catalytic components in most catalysts that have been developed and studied for chemical reactions. These catalytic components often are complementary and beneficial to the active catalytic component needed for a reaction. The following processes are commonly used to prepare multicomponent catalyst materials:

• A simple physical mixture of different catalyst components.¹¹ 

• Composite materials with active catalyst components and/or promoted (attached, doped, loaded, and supported) by other catalytic/non-catalytic components deposited onto a support material. The support material may provide a means to only disperse the active phase or could also participate in the reaction. Additionally, the proximity of the support and supported metals can produce Strong Metal–Support Interaction (SMSI) effects, which can significantly impact the activation barrier of a catalytic process.^12–15 

• Matrix materials (polymer, carbon, etc.) modified with metals, metal oxides/hydroxides, organic motifs, and/or functional groups (such as sulfonic acid groups, nitrate/nitrite, and hydroxide groups). Unlike the aforementioned supported materials, the SMSI effects are often not observed, but having high surface area and functional groups of the matrix significantly improves the catalytic activity and selectivity towards desired products.^16,17 

• In situ developed mixed metal oxide catalyst materials (such as CuNiCoO)¹⁸  and/or supported materials (such as ZnO/graphene oxide),¹⁹  prepared by sol–gel hydrolysis or co-precipitation followed by controlled heat treatment, are also an integral part of the industrial catalytic processes which show high stability of its catalytically active phases despite harsh reaction conditions and long operation hours.²⁰ 

• Hierarchically organized materials (specifically zeolites) consisting of micropores, mesopores, and macropores are also popular in many industrial processes. The presence of added porosity level in these catalysts/materials allows the reactants and products to readily enter and leave the catalytic sites, which, in turn, maximizes the utilization of the active sites, facilitates the diffusion-constrained reactions, and helps improve reaction selectivities by allowing certain reactants to pass though the tunable pores with catalytic sites.^21,22 

• Finally, we should also discuss the materials obtained from overgrowth, and co-growth of crystallites, where different crystal growth processes form different structures and expose different facets of the active sites. For example, a core–shell structured material is obtained through the overgrowth of the shell composition (usually metal oxide) onto the core composition (usually metal),²³  whereas an alloyed nanoparticle is obtained by the co-growth of two or more metal crystallites.²⁴  A great control over these can significantly advance the reaction rates and selectivities.²⁵ 

Here, we have enlisted the ways different components of a composite material can contribute to its overall catalytic performance.

• Developing a simple supporting effect by providing a high surface area to the catalytic components and maximizing its exposure to the reactants. Such as, the formation of V₂O₅ domains on different (Al₂O₃, ZrO₂, TiO₂, and CeO₂) supports increases the number of active sites exposed on the surface.²⁶ 

• Influencing the electronic nature of the catalyst surface, which, in turn, can lower activation energies and increase the activity and selectivity of the active components.²⁷  This can be done by systematically tuning the active site compositions on catalyst surface and interaction between the active site and the support.²⁸ 

• Stabilizing the nano–microstructures and/or active components of the catalyst by non-active sites is of great importance as well in industrial process developments as it helps lower the catalyst regeneration cost. Several types of low-energy interfaces, such as twin boundaries, high and low-angle grain boundaries, and interphase boundaries are effective in stabilizing catalyst nanostructures.²⁹ 

• Forming new compounds or sites which act as active catalytic components or catalyst stabilizers. As observed for Pd–Zn/Sibunit catalysts during the thermal treatment in hydrogen as used in methanol synthesis and alkyne hydrogenation reactions.³⁰ 

• Creating two or more catalytic sites where they can influence the catalytic activity together or independently. Such as the Cu/Co sites in MOF-74 which co-catalyze the aerobic oxidation of styrene.^31,32 

• Creating two or more catalytic sites which work differently under varying conditions. This makes these catalysts useful for various catalytic processes. For example, cobalt and manganese sites were introduced to this cobalt/manganese oxide catalyst to work differently. While manganese oxide favored the dehydrogenation reaction, cobalt oxide assisted the hydrogenation kinetics.³³ 

• Controlling the acid and base site densities and strengths and site density can also influence catalytic performances of acid–base catalyzed reactions, which was found to play a major role in K₃PO₄/NaX zeolite catalysts for controlling the alkylation kinetics of toluene with methanol.³⁴ 

• Improving the adsorption, desorption, and diffusion rate of reactant molecules through porous (containing micropores and mesopores) structures which influence the catalytic performance of a composite material significantly. As observed in materials containing multiple active sites such as La_0.6Sr_0.4Co_0.2Fe_0.8O_3−δ, because of the varied absorptivity, and diffusivity of different elements towards O₂.³⁵ 

In ancient times, people in Europe, Australia, and North America used to use paints based on iron oxide, chalk, or charcoals to paint rock walls.³⁶  This was a primitive form of coating, which was later modified to protect and decorate various materials used in society. With time, its scope broadened and technologies advanced so much that nowadays we apply coatings on various objects ranging from macroscopic to microscopic dimensions. While coatings on substrates of macroscopic dimensions are often used to protect from environmental factors such as corrosion, and/or to enhance their integrity (such as ceramic coatings are used to protect cars from minor scratches), coatings on substrates of microscopic dimensions are often used to modify the surface properties of a material, such as its physiochemical properties, and/or to introduce new functions. Among various applications of microscopic coating, it has been applied to preparing various composite materials for heterogeneous catalysis applications to modify and decorate catalytic materials for different reactions. The coatings have been used to not only prepare catalysts on macroscopic support structures (e.g., monoliths) but are also catalytic materials (powders) to achieve specific interactions between active components on the surface to enhance stability and reduce deactivation during reactions.³⁷  Catalysts, in general, have major industrial implications, which in turn, makes the use of coating processes critical. In this regard, advanced coating processes that are industrially friendly (i.e., inexpensive, efficient, and use non-toxic materials) are in high demand.

To date, thousands of coating systems and procedures have been developed, but in most cases, they fail to fulfill (i) the industrial demands of producing high-performance coating at a relatively low cost,³⁸  or (ii) the environmental demands of minimizing hazardous wastes from coating processes.³⁹  This leaves only a few procedures that are used widely to coat various substrates, including heterogeneous catalysts, using liquid, gas, or solid precursors. In the following sections, those processes have been described briefly in the (i) conventional coating methods and (ii) advanced coating methods sections.

In this section, we have described the processes that are well-known and are widely used across many industrial sectors for various applications.

The vapor deposition method is a process in which chemical vapors are used to coat a substrate with thickness in the range between nanometers to micrometers. This coating technique has been widely used for many years to produce different types of materials, including fibers, nanotubes, powders, thin films, and multilayer coatings.⁴⁰ 

CVD is one of the most common processes which is used to coat metallic and/or ceramic compounds. In this process, a solid material is deposited, in the form of coating, from a gaseous phase on a substrate and can provide a coating of 5–10 μm. This is achieved through a chemical reaction that occurs when the precursor vapor passes over the heated substrate surface. Depending on the process, reaction conditions, precursor, and substrate used, CVD can be of several types such as atmospheric pressure chemical vapor deposition, metal–organic chemical vapor deposition, low-pressure chemical vapor deposition, laser chemical vapor deposition, photochemical vapor deposition, chemical vapor infiltration, chemical beam epitaxy, plasma-assisted chemical vapor deposition, and plasma-enhanced chemical vapor deposition.⁴¹  This process can be applied to a wide variety of materials to coat precision surfaces and the materials produced by this process can withstand extreme temperature variations. But unfortunately, this process cannot be performed at low temperatures. This process typically requires temperatures in the range of 450–1050 °C.⁴²  To date, this method has been used to prepare various supported monometallic catalysts such as Mo/Al₂O₃, Pd/SiO₂, and Pt/zeolite.⁴² 

Similar to CVD, PVD is also a vapor coating technique. In PVD solid materials are used as the coating precursor, unlike gaseous precursors used in CVD, to coat a substrate with a thickness of 0.5–5 μm. The absence of any solvent or reducing agent makes it a cleaner coating process requiring less post-deposition treatment to clean the surface after deposition. Here the coating process involves four main steps: (i) vaporization of the material to be deposited by a high energy source; (ii) transportation of the disintegrated material to the substrate to be coated; (iii) reaction between the disintegrated atoms and the reactive gas during the transportation stage; and (iv) deposition of the coating material on the substrate surface. Depending on the energy source used to evaporate the coating materials, PVD can be majorly divided into three types (i) thermal evaporation PVD, (ii) sputter deposition pvd, and (iii) arc vapor deposition PVD.⁴¹  These processes typically operate under temperatures in the range of 250–450 °C and are used to prepare various supported metal, metal carbide, metal sulfide, and metal nitride catalysts.^43–45 

The sol–gel coating process involves the formation of (i) sol (monomers suspended in a colloidal media) and gel (semisolid particles suspended in a colloidal media); (ii) application of gel at the substrate; and (iii) formation of coating material. This process is favored in preparing various heterogeneous catalysts due to its high chemical flexibility, and tolerability of the catalyst surface.⁴⁶  This process is widely used to prepare monolith catalyst structures, as observed in preparing Ni-based monolith catalyst for partial oxidation of methane.⁴⁷ 

In this age-old process, a small drop of the coating material is added onto the center of a substrate, which spins around an axis perpendicular to the coating area at a controlled high speed to form thin films/coating layers. In short, it is comprised of the following stages: (i) deposition, (ii) spin-up, (iii) spin-off, and (iv) evaporation. The thickness of the films/coating layers obtained in this process depends on the parameters of coating solutions (such as viscosity, drying rate, percent solids, and surface tension) and process (such as the rotation speed and time).^48,49  This process is also widely used to prepare catalytic membrane fabrication, such as porous TiO₂/α-alumina membrane for methylene blue degradation under sunlight.⁵⁰ 

Electro- and electroless chemical plating, are processes in which a thin conductive metallic film is deposited by the reduction of a metal ion from solution onto a conductive surface in an electrochemical cell. In this process, the substrate to be coated is used as the cathode (where reduction or metal deposition happens), the solution that carries/transports the charged ions (of the metal to be coated) is called the electrolyte, and the metal block (of the metal to be coated) or the conduction material of the electrochemical cell where oxidation happens is called the anode. This surface modification method is desirable in preparing various catalysts due to its simplicity, cost-effectiveness, and wide scope. Depending on the source of electrons, the process is either called (i) electrochemical plating or (ii) electroless plating, which are described in detail in the following sections.

Electrochemical plating, also known as electroplating and electrochemical deposition, is a process in which the electrons required to reduce a metal ion in solution are supplied from an external source in the form of electricity. During this process, a chemical surface treatment is often required to prevent oxide formation on the anodes (metal blocks of the metal to be coated), and finding the appropriate chemical surface treatment process for a given metal oxide is one of the major challenges involved in plating processing.⁵¹  In addition, the surface in-homogeneity of the substrate and the uneven distribution of current density in the plating bath also create further problems in the plating process resulting in non-uniform coatings.⁵²  Despite these disadvantages, this process is highly exciting and useful for preparing various catalysts and decorating their surfaces.^53,54  The use of this simple, efficient, and energy-conserving procedure in depositing a wide variety of metals and alloys in the form of a thin coating makes this process unique and advantageous over others.

Electroless plating, also known as autocatalytic plating, is a chemical process in which the reducing electrons are supplied by a chemical reducing agent in the solution or from the material itself and uses no electricity. This process is capable of co-depositing catalyst substrates with second-phase particles such as alumina, carbides, and carbons, and fabricating catalyst surfaces according to the need. More importantly, this process does not suffer from the same disadvantages as those noted previously for electroplating, such as non-uniform coating, complex set up, and challenging scale up.⁵⁵  This makes this process a great alternative for industries that seek a cost-effective, simplified method for coating their catalysts,^56,57  as shown by Fujii et al.,⁵⁸  in preparing Pt particles for polymer electrolyte fuel cells.

During the electrochemical plating process mentioned previously, another process occurs when the oxide ions (if present in the electrolyte) react with the metal ions on the anode. The process is called anodization and it is an electrolytic passivation (oxidation) process that converts metal and alloy surfaces into corresponding decorative and durable oxide surfaces. Although this process is primarily known for making materials corrosion-resistant and paint adhesion to the substrate, more recently it has been used to decorate catalyst surfaces to introduce unique properties (hardness, porosity, thickness, and corrosion resistance)⁵⁹  and catalytic components as shown by Touni et al.,⁶⁰  in preparing IrO₂/Ir(Ni) film for electrocatalytic oxygen evolution reaction. Conventional anodizing process includes six major stages: (i) mechanical treatment; (ii) degreasing, cleaning, and pickling; (iii) electropolishing; (iv) anodizing using AC or DC current; (v) dyeing or post-treatment; and (vi) sealing.⁶¹ 

While the quality of the anodized film depends majorly on the electrolyte and the experimental conditions, electrochemical inhomogeneity of the coating substrate (i.e., the presence of uneven current density, which occurs on an electrode surface due to uneven polishing, materials defects, or irregular chemical adsorption and leads to two or more simultaneous electrochemical reactions and/or unstable reaction rates) and localized heating are minimized to improve the coating uniformity.

Laser irradiation, also known as laser ablation or photoablation, is the process of removing materials from a solid surface by irradiating it with a pulsed laser beam. The amount of material to be removed by a single laser pulse depends on the depth over which the laser energy will be absorbed, which, in turn, depends on the optical property of the substrate, laser wavelength, laser flux, intensity, and irradiation time. This highly controlled and precise technique is extensively used in spectroscopic and bio-medical applications and is not a typical coating process but more a technique to create catalyst surfaces or catalytic components in layered materials, as shown by Cao et al.,⁶²  in creating an oxygen-vacancy-rich zirconium oxide layer on the Zr-sheet surface, and Peng et al.,⁶³  in embedding a single atom Pt catalyst on graphene support, respectively.⁶⁴  Fine control over localized temperatures and heating times are the key attributes that distinguish this technique from the traditional reduction processes.⁶³ 

SPS is used to deposit coating materials of submicron or nanometer sized particles. The suspension used in this technique contain fine particles of coating material that are dispersed in a liquid phase and are injected into the plasma as a liquid stream. A typical feature of SPS coating are the columnar structures, i.e. the vertical cracks across the thickness of the coatings. This allows the coated materials to maintain high surface areas, and as a result this process has gained significant attention of the researchers in preparing catalyst structures, as used in developing various carbon supported catalysts (such as Ni–Co–Fe/C) for Fischer–Tropsch synthesis.⁶⁵  The thermal conductivity of this technique strongly depend on coating features, and therefore, to control the coating features well and achieve a desirable coating microstructure, injector properties and spray parameters needs to be accurately optimized.⁶⁶ 

ALD, a variant and advanced version of CVD, is an ultra-thin-film (of a few nanometers) deposition technique based on a gas-phase chemical process that provides excellent uniformity and atomic precision. This is a self-limiting process in which the precursors or reactants react with the substrate surface sequentially, resulting in the formation of ultra-thin films (per cycle) with excellent growth control. A typical ALD process involves a cycle of four key steps which can be repeated as many times as necessary to achieve the required film/coating thickness: (i) dosing of the substrate with a precursor vapor that adsorbs on and reacts with the substrate surface. (ii) Purging of all residual precursors and reaction products, (iii) controlled exposure of the surface with reactive radicals under controlled temperature to oxidize the surface and remove surface ligands, and (iv) purging of all residual precursors and reaction products. Depending on the source of temperature, presence of a catalyst, and types of reactants and substrates used for this process, it can be of many types such as (i) thermal ALD, (ii) plasma ALD, (iii) photo-assisted ALD, (iv) metal ALD, (v) particle ALD, (vi) polymer ALD, and (vii) catalytic ALD. Although this process is known for fabricating microelectronics and semiconductor devices, there is a growing interest in using this atomically precise technique in designing and preparing single-atom heterogeneous catalysts for various applications as discussed by Fonseca et al.^67,68 

In addition to excellent control over the layer thickness, this process also offers a superior ability to penetrate high-surface-area porous substrates, surface chemical selectivity, and industrial scalability, which makes this advanced coating process useful in many research and industrial applications, specifically catalysis.⁶⁸ 

All of the previously mentioned advanced coating processes are critical to developing the next-generation catalysts for a wide range of applications, including chemical synthesis, and environmental remediation.⁶⁹  In addition, these processes are being practiced industrially for various other applications, such as ALD which is used in both catalyst preparation for propane dehydrogenation processes,⁷⁰  and biomedical applications including the development of flexible sensors, nano-porous membranes, and thin biocompatible coatings.^71,72  Similarly, electro/electroless chemical plating processes are used in developing catalyst materials for energy applications,⁷³  and developing corrosion and wear-resistant materials.⁷⁴  These applications make the advanced coating processes extremely interesting to researchers across industry and academia due to their superior control and precision at the atomic level in designing heterogeneous catalysts. While these advanced coating techniques are creeping into industrial practices, it is also essential to couple them with renewable energy resources to minimize the carbon footprints associated with them otherwise.

In this regard, the aspect of catalyst discovery by AI and ML methods are discussed, which can be programmed based on existing data sets from peer-reviewed literature, and are anticipated to help scientists determine catalyst compositions, catalytic conditions, and reaction mechanisms at lesser time and cost. We have described in the following section how some of the AI and ML methods are being used to discover these aspects of catalysis and contributing to basic catalysis research.

The rapid evolution of computing ability, in terms of storing and processing an enormous amount of data, has significantly developed the fields of AI over the past decade. This has, in turn, dramatically advanced the fields of language processing, image recognition, and automation. In this regard, an important tool (of AI) known as ML, which uses algorithms to learn from data, has evolved to detect patterns and make fast and accurate predictions in applications with a large parameter space.⁷⁵ 

Though ML was first reported in the 1940s, it was not widely adopted as a practical technology until recently.⁷⁶  Despite its growing influence in various domains, its application in catalysis is still at its infancy. For example, conventional catalyst development involves designing and synthesis of catalysts by an iterative trial-and-error method, which is a time-consuming and high-cost process. Automated ML processes requiring low computational cost, based on the latest algorithms and theories, could be used to parse through vast and complicated catalytic data sets of material properties, computational predictions, and process variables to minimize associated time and cost and assists in predicting novel catalytic designs, compositions, and catalytic mechanisms.^77,78  In the following sections, we discuss current applications that are aided by using ML in the field of catalysis.

As an interdisciplinary field of study involving computer science, statistics, and various subjects in data science, ML can be classified, in many ways, based on the particular task employed for solving problems. However, the classifications based on the method of learning and algorithms (a set of rules that are followed to do specific tasks or calculations by a computer) used are the most popular. The learning method can be distinguished as supervised and unsupervised learning. While the former is the most widely used and has made much of the practical successes in deep learning and discovering or predicting models, the latter works well when dealing with big data sets to find hidden patterns in data that have not been labeled with identifying characteristics, properties, or classifications. The two most common techniques used in unsupervised learning are clustering and K-means. Whereas, supervised learning can be categorized into many types, among which multiple linear regression analysis (MLRA), artificial neural networks (ANNs), K nearest-neighbor (KNN), and random forest (RF) are the most popular types.

The simplest ML method, MLRA, models any property of an object by the linear combination of its descriptors. Whereas the most popular algorithm, ANN, mimics the working principle of biological neurons found in humans and is widely used in both regression and classification analysis. The RF, on the other hand, is based on the construction of a training set and a predictor tree that maps out the observation of a given variable through different branches and nodes to reach an accurate and stable prediction.

However, it is to be noted that each ML method has its pros and cons, which makes it critical to select them meticulously depending on the analysis need and the size of the database. For example, in the case of a small data set (that has low complexity or high bias), linear regression and classification methods can be employed, whereas, for a big data set (refers to data storage amounts in excesses of one terabyte), nonlinear methods such as ANN and KNN methods are preferable.

In recent years, these ML methods are making their contributions to different branches of chemistry. Its use in determining catalyst compositions, properties, and reactivities is significantly advancing the fields of catalysis which are described in the following sections.

Heterogeneous catalysis involves interactions between a chemical molecule with a catalyst substrate that has a different physical state. The ability to vary reaction conditions (temperature, partial pressure, reactants, and space velocity) during a reaction, catalyst properties (composition and weight), and synthesis/pre-treatment conditions make it easier to generate large data sets in heterogeneous catalysis. This makes the heterogeneously catalyzed processes data abundant,⁷⁹  and in multiple instances AI/ML have been successfully used to design, discover, and screen catalyst materials.^80,81  For example, in one of the earlier works, Sasaki et al.⁸²  successfully employed an artificial neural network based model to estimate catalytic activities of Cu/ZSM-5 for nitrogen monoxide (NO) decomposition reaction and further validated their findings by experimental results.

In this regard, it must be mentioned that quantum mechanical (QM) calculations have significantly contributed in the past few decades to the discovery of catalyst compositions in systems such as bimetallic alloys where modeling diverse active sites is a topic of discussion.⁸³  Although these high-level quantum-mechanical calculations can accurately obtain reactivity descriptors, their expensive computational cost plays a limiting factor. ML methods, on the contrary, can accurately model the reactivity of catalysts and identify their descriptors using significantly fewer computational resources and time. For example, for each stable low-index facet of a bimetallic crystal, hundreds of possible active sites exist, and to systematically search them, quantum-mechanical calculations may take hours and days, while ML-based neural network potentials can model those surfaces in a relatively very short time.⁸⁴  To date, this has been followed in discovering other important catalytic factors responsible for various reactions such as the decomposition of gas molecules on metal nanoparticles, intermetallic bond distances in determining the HER activity, and activation energy of catalytic reactions by ML methods and algorithms like Bayesian linear regression, regularized random forest, and support vector machine, respectively.

Similarly, using the descriptors (such as density and the enthalpy of fusion) of the active metal sites, ML methods (such as gradient boosting regression (GBR)) can accurately predict their d-band centres (together with sp-bands, d-bands, and electronegativity), which in turn, are widely used to predict activity trends of metal catalysts and their alloys.^85,86  Depending on the nature of the catalytic sites, other finger-print features are used to predict the best catalyst composition for specific reactions, such as rotational angle, electronegativity/electron-affinity, and geometrical structure (along with bonding characteristics). These serve as effective input features to predict structures of novel transition metal dichalcogenides (for superior water splitting performance);⁸⁷  bimetallic alloys (for superior CO₂ reduction performance);⁸⁸  and B-doped single atoms (for superior N₂ reduction performance),⁸⁹  respectively,⁸⁶  A summary of various ML methods, their classifications, and applications in catalysis can be found in Table 1.

In addition to these reaction properties, reaction conditions also play a critical role in productivity optimization. In an early work, the backpropagation ANN model was used to study and analyze the correlation between variables that express NO decomposition reaction on Cu/ZSM-5 catalyst.⁸²  In this regard, the Gaussian process surrogate model ML and black-box Bayesian optimization can also be used to obtain better reaction outcomes, avoiding expensive high-throughput experimentation to treat them algorithmically and predict solvent compositions and optimal reaction temperatures, respectively.

Despite the applied contributions of ML having increased in recent years to catalytic studies, ML can still be considered in its budding stage in the field with many underexplored research opportunities. In recent work, Perez-Ramírez et al.⁹⁰  have discussed the development of a ML framework and its challenges in predicting catalyst performance (space-time yield, STY) from experimental descriptors for CO₂ hydrogenation to methanol. This study not only identifies the major factors contributing to the STY of a catalyst but also denotes how this model can be used to train other types of materials or even descriptors (such as structural or electronic) of a material. These studies indicate that AI/ML will be an integral part of our research in the coming years and will influence the design and development of heterogeneous catalysts and may even precisely predict their reaction conditions. It may even bring a revolution in chemical industries by lowering the costs associated with every aspect of catalyst discovery from formulation to synthesis to evaluation.

For small systems, catalyst powders can typically be assessed in packed beds when a plug flow behavior is desired or when the reaction is mass-transfer limited,¹⁰  giving high selectivity and fewer byproducts. However, for larger systems, mass and heat transfer, and large pressure drop becomes an issue that requires improved geometries. Thus, these challenges can be addressed by shaping the catalyst into different forms³  such as pellets, granules, extrudates, monoliths, and foams (see Fig. 1).

Amongst all the shapes, monoliths are popular because of high void fractions,¹⁰  and interconnected or separated channels that allow for linear axial flow, low-pressure drop, lightweight, and customizable size and shape.⁹²  Additionally, monoliths can be made of different materials such as metals or ceramic that can offer other advantages such as removing exothermic heat (metallic monoliths) or maintaining heat as an adiabatic reaction (ceramic monoliths).¹⁰  Conventional processes such as extrusion pose limitations on the achievable complexity in monolithic structures and on the choice of materials. This is where 3D printing and AM can play a big role by generating complex structures of smaller dimensions having unique and tortuous geometries, and high surface areas to enhance gas contact.

As discussed previously, the typical supports used in catalytic applications are metal oxides covering ceramics, zeolites, and various other porous materials such as MOFs as they provide high surface area. A variety of these porous support materials have been 3D-printed in monolith form using different techniques to achieve tunable channel distribution, channel design, and stacking and these materials can be broadly classified as polymeric materials, ceramic materials, metallic materials, and carbon materials.⁷  Out of these, the application of polymeric materials is mainly involved in the printing of the reactionware (chemical reactor components) and sometimes may include the use of a dopant (e.g., graphene or diamond)⁷  to enhance the properties of the base polymer and to add additional properties. Occasionally, for some low-temperature reactions (<250–300 °C), polymer-based 3D-printed catalytic supports can also be used to load active material onto.⁹³  Polymer-based 3D-printed supports or reactionware are often catalytic, but they are much easier to print using a wide array of techniques. Whereas, ceramic, metallic, and carbon materials can be used as reaction ware, catalyst, or catalytic support. A summary on the variety of monolithic supports that have been 3D-printed and applied in different catalytic applications can be found in Table 2.

As can be observed from Table 2, the most commonly used methods for 3D printing supports include DIW or robocasting, inkjet printing, and occasionally SLP. DIW offers the best customization because the ink can be varied quite easily, and a variety of materials can be 3D-printed. A short summary on these different methods can be found in Table 3.

Monoliths for a chemical reaction can be designed in different configurations and channel orientations depending on the need of the application. Laguna et al.⁷  identified three primary configurations that are mainly 3D-printed: (i) woodpile configuration, (ii) monolith with channels, and (iii) isoreticular foams. The woodpile configuration is the most commonly used because of the simplicity, tunable density of cavities/channels, tunable degree of stacking, and comparable/superior performance against traditionally manufactured monoliths by extrusion. Woodpile-based structures exist in two basic variations, straight and staggered configurations, and can often be obtained in cylindrical or cubic shapes. On the other hand, isorecticular foams and monoliths with channels are known as periodic open cell structures (POCS) and are not frequently used. All these structures are depicted as shown in Fig. 2.

Several examples of applications of monoliths with woodpile structure exist and some of them are already summarized in Table 2. The next sections will discuss different types of supports that have been 3D-printed.

Many of the heterogeneous catalytic reactions are carried out at high temperatures (700–1000 °C), e.g., steam reforming of methane for producing H₂.¹²⁵  For withstanding such harsh conditions, catalysts need stable base support like ceramics and various ceramic supports such as Al₂O₃, SiO₂, TiO₂, ZrO₂, and mixed metal oxides are widely used. It is imperative that 3D printing of these materials will facilitate the development of structured ceramic catalysts.

Among the various supports printed, Al₂O₃ based catalysts have been the most prominent. One of the earliest reports of a 3D-printed material was Cu/Al₂O₃ as a catalyst support by Tubio et al.¹²⁶  The authors prepared an ink made up of Al₂O₃ mixed in an aqueous Cu(NO₃)₂ · 2.5H₂O solution along with viscosity modifiers and a cationic polyelectrolyte. This ink was 3D-printed using the DIW technique. The catalytic activity of these 3D-printed catalysts was tested in Ullmann reactions, and they provided excellent yields (78–94%) to the desired products. Additionally, the catalyst did not show any signs of leaching out confirming the stability. In the same group, Azuaje et al.¹⁰⁰  used 3D-printed Al₂O₃ for Biginelli and Hantzsch reactions. The authors prepared alpha-alumina by robocasting colloidal alumina ink into a straight channeled woodpile monolith structure. This provided a high surface-to-volume ratio and a controlled porosity for the catalytic structure resulting in high throughput synthesis of 1,4-dihydropyridines and 3,4-dihyropyrimidin-2(1H)-ones. Similar to this work, Azuaje et al.¹⁰²  prepared a graphene oxide–alumina (GO–Al₂O₃) ink that was used to 3D print a monolith with the DIW technique. This GO–Al₂O₃ catalyst was used for Paal–Knorr synthesis of pyrroles (yields 85–93%) and for assembly of benzimidazoles (yields up to 92%). This catalyst was also found to be highly recyclable when tested in 10-cycle repeatability.

Middelkoop et al.⁹⁸  used Ni to support on 3D-printed Al₂O₃ for CO₂ methanation reactions. They used the direct ink writing method to 3D print Al₂O₃ from colloidal ink. Very high conversion (∼96%) and selectivity to methane (∼95%) was obtained using this catalyst. Wang et al.⁹⁹  used a different technology (DLP) to 3D print gamma-alumina catalysts. This catalyst provided high conversions (∼99%) for methanol to hydrogen reaction as compared to standard spherical catalyst (not 3D-printed) which showed no activity.

Another widely 3D-printed support includes SiO₂. Notable works include two articles from Diaz-Marta et al.^93,104  In the earlier work,¹⁰⁴  the authors 3D-printed supports from SiO₂ ink using robocasting. This support was functionalized with Cu and Pd compounds to obtain Pd and Cu devices (monolith system). These monolithic devices were validated for catalytic efficiency in copper alkyne–azide cycloaddition (CuAAC), Sonogashira, Stille, and Suzuki reactions. These 3D monoliths showed high mechanical strengths, no metal leaching, and easy catalyst recyclability for multi-catalytic multicomponent reactions. In the other work,⁹³  for a similar reaction, they demonstrated a tri-catalytic system based on immobilized and compartmentalized metal species in supports. This involved Cu²⁺ in a 3D-printed capsule, Cu⁺ in nano-particle form, and Pd⁰ coated on SiO₂ monolith. This tri-catalytic system demonstrated that the two different 3D printing techniques (FDM and DIW) can be combined to manufacture customized catalytic devices allowing easier catalyst separation and reusability.

Other ceramic supports that are 3D-printed include CeO₂. One of the earliest attempts includes the work from Leclerc and Gudgila,⁹⁴  who prepared the CeO₂ monoliths using the Robocasting method. Active Rh metal was then deposited on this monolith using an impregnation method. This catalyst was compared to other similar ones such as Rh supported on an Al₂O₃ monolith, Rh/Ce supported on an Al₂O₃ monolith, and Rh supported on CeO₂/Al₂O₃ (1 : 1 wt) monolith. These catalysts were subsequently tested for partial oxidation of methane and observed that the addition of CeO₂ promoted higher methane conversion (0.94 vs. 0.9) and water gas shift activity (high H₂ selectivity, 0.95 vs. 0.87; low CO selectivity, 0.83 vs. 0.85) as compared to only Al₂O₃ supported catalyst. Lucentini et al. ⁹⁵  3D-printed CeO₂ (using material extrusion) to support Ni active metal and tested it for ammonia decomposition reaction. The 3D-printed CeO₂-supported catalysts were found superior as compared to conventional cordierite honeycomb supported catalysts and comparable to the powder-based catalysts. In a subsequent study,⁹⁶  the same group deposited Ni–Ru on 3D-printed CeO₂ support. They observed that the experiment and the model results converged very well for ammonia decomposition reactions. This is significant because the 3D-printed supports could match the activity estimated from theoretical calculations.

TiO₂ support has also been 3D-printed with different technologies such as additive superposition and FDM. Elkoro et al.¹⁰⁷  prepared TiO₂ monoliths by additive superposition of microfilaments of TiO₂ paste. These monoliths were then doped with Au nanoparticles (NPs) and tested for hydrogen photoproduction. The pre-impregnated Au/TiO₂ catalysts (Au NP impregnation in TiO₂ before 3D printing) showed a homogeneous Au NP distribution whereas post-impregnated catalysts (after 3D printing) showed an asymmetric distribution with high Au NP concentration at the surface and lower at the inner regions. This high surface concentration of Au NP in the post-impregnated catalyst exhibited enhanced catalytic performance even at lower Au loadings. The diameter of the filaments also inversely influenced the rate of H₂ photoproduction with the best activity of 0.24 mol H₂ min⁻¹ g_Au⁻¹ obtained with microfilaments of 200 µm diameter. Sevastaki et al.¹⁰⁸  3D-printed fully recycled TiO₂–polystyrene nanocomposite photocatalysts for use against drug residues. In this work, authors first made a slurry out of recycled polystyrene (PS) to which TiO₂ NPs were added and the suspensions obtained were converted into a dense precipitate, which upon drying gave a nanocomposite solution that was converted into a filament and then extruded using an FDM printer. This catalyst showed promising photocatalytic properties reaching an efficiency of 60% after three cycles of reuse in 200 ppm of APAP aqueous solution under UV-A irradiation.

SiC is a well-known tough ceramic that is widely used in catalysis applications as well. One of the recent works from Quintanilla et al.¹²⁷  reports 3D printing of Fe–SiC based catalyst with robocasting/DIW technique. In this, Fe–SiC ink was prepared by mixing organics such as dispersing and viscosifying agents (6 wt%), Fe/SiC (∼56 wt%), and DI water (∼38 wt%). The monoliths showed excellent mechanical strength, high catalytic activity, efficiency in the H₂O₂ decomposition, and good long-term stability (∼350 h). Up to 80% phenol conversion was obtained (at low space velocities and 75 °C) in the wet peroxide oxidation process with the 3D-printed SiC monoliths and treated at 1200 °C.

ZrO₂ supports do not appear that popular in catalytic applications as hardly any studies on 3D printing of it for catalysis can be seen. However, it has been 3D-printed for several other applications with the most popular being printing of dental parts. All of these different applications are covered in a comprehensive review by Zhang et al.¹²⁸ 

Among the SiO₂–Al₂O₃ based supports, zeolite ZSM-5 has been widely 3D-printed by the Rownaghi group using FDM technology. In one of the early studies,¹¹³  they 3D-printed H-ZSM-5 catalysts in monolith shape to improve scalability and performance. When compared to the powder counterparts, 3D-printed ZSM-5 were found to be more stable for n-hexane cracking and more selective to light olefins. In another study,¹⁰⁹  the group 3D-printed ZSM-5 and doped it with different metals such as Ga, Cr, Cu, Zn, Mo, and Y. These catalysts were further studied for methanol to hydrocarbons (MTH) reaction with and without CO₂ addition. The metal dopants altered the physicochemical properties of ZSM-5 and varied the product distribution. The surface area dropped considerably (14–23%) whereas mesopore volume dropped by (30–47%) depending on metals. Also, the acidity (number of acid sites) of metal doped catalysts had slightly increased compared to the base zeolite due to presence of metal oxides. Under the reaction atmosphere, most metal dopants showed preference to light olefins whereas with CO₂ addition to the feedstock, benzene, toluene, and xylene (BTX) selectivity was significantly increased especially for Y and Zn doped ZSM-5 catalysts. In another study,¹¹⁰  the group studied Cr-, Cu-, Ni-, and Y-doped 3D-printed ZSM-5 monolith catalysts for n-hexane cracking. The metal dopants were directly added in the form of metal nitrates to the printing ink solution thus allowing to directly print the metal-doped ZSM-5 catalysts. The addition of metal dopants influenced the acidity and porosity of these catalysts. Cr, Cu, and Ni doping favored the selectivity toward BTX, whereas Y doping favored the selectivity toward light olefins. In another study,¹¹²  the same group 3D-printed HZSM-5 with FDM technology, and using a secondary growth method they added SiO₂ and SAPO-34 to tune mesopores and the acidity of the catalysts. These catalysts were further studied for methanol to olefins (MTO) reaction and were found to suppress coke formation because the reduction in Brönsted acid sites hindered the formation of paraffins and aromatics.

Cepollaro et al.¹²⁹  synthesized 3D-printed geopolymer (GP) based ZSM-5 catalysts with DIW method. In this work, GP catalysts with up to 37 wt% content of ZSM-5 were prepared and further characterized using XRD, SEM, BET, and H₂ TPR. It was observed that the acid pre-treatment of GP/ZSM-5 before Cu exchange helped in forming exchangeable Cu positions without affecting the crystalline structure and porosity. This catalyst with exchangeable Cu positions in GP/ZSM-5 was found to be active for selective catalytic reduction (SCR) of NO_x as seen in another work¹¹¹  from the same group. In this, 3D-printed GP–ZSM-5 based catalysts were prepared by adding ZSM-5 as a filler during the ink preparation and the content was increased up to 60 wt% (dry basis). The active metal Cu was introduced by ion exchange. These GP-Cu/ZSM-5 catalysts were used for NH₃-SCR reaction and provided a high NO conversion and selectivity to N₂ over a wide range of temperatures.

Among metallic monoliths, stainless steel (SS) based ones have been primarily 3D-printed. Agueniou et al.¹¹⁷  3D-printed 15-5PH SS honeycomb monoliths with direct metal laser sintering method (DMLS) and wash-coated them with Ni/CeO₂–ZrO₂ powders. These catalysts were further tested for the dry reforming of methane (DRM) reaction and compared against the catalyst powder deposited on a conventional cordierite monolith. At 750 °C and 800 °C, for the same catalytic loadings, 3D-printed SS-based catalyst showed lower conversion and slightly lower H₂/CO ratio as compared to cordierite-based catalyst. However, at 900 °C, conversions and H₂/CO ratio were both higher than cordierite and there was no apparent deactivation. This was attributed to the initial nickel content in the SS monolith (when tested without any washcoat, it showed ∼50% conversions at 900 °C). Also, there was no activation time in the case of SS that indicated enhanced heat transfer in the metallic monolith compared to cordierite. Danaci et al.¹¹⁸  3D-printed SS-316L monoliths using the robocasting method and coated them with Ni/Al₂O₃ catalyst. This catalyst was then tested for CO₂ methanation and compared against a conventional powdered Ni/Al₂O₃ catalyst. At higher temperatures (350–450 °C), higher CO₂ conversions were observed for monolithic catalysts when compared to powder catalysts. In particular, the staggered catalyst showed higher conversions than conventional powder for the whole temperature range (250–450 °C). All of this was mainly attributed to the improved heat and mass transfer in the monoliths. In a subsequent SS-based study, Danaci et al.¹²¹  3D-printed additional Cu-based monolith supports and deposited Ni/Al₂O₃ catalysts onto it. These 3D-printed Cu- and SS-supported Ni/Al₂O₃ catalysts were evaluated for CO₂ methanation reaction and compared against powder Ni/Al₂O₃ catalysts. The catalysts with 3D-printed supports showed slightly higher methane productivity (3–4 mmol g_cat⁻¹ h⁻¹) than powder-based catalysts (0.7–3 mmol g_cat⁻¹ h⁻¹) in 350–450 °C range. These Cu- and SS-supported catalysts were further evaluated in a scaled-up study¹²⁰  where a reactor fifteen times larger than a lab reactor was used. A three times higher methane productivity (256 mmol g_Ni⁻¹) than lab-scale, high stability for 80 h, low amount of carbon formation, and no hot-spot was observed in the larger reactor setup.

Lefevere et al.¹²²  3D-printed SS-316L supports using 3DFD (three-dimensional fiber deposition) method and washcoated them with a ZSM-5 catalyst. The washcoating method was optimized and the catalysts were evaluated for methanol to light olefins reaction. Additionally, 3DFD-based catalysts were compared with the cordierite honeycomb based catalysts and also with a packed bed of pelletized ZSM-5. It was observed that a 3DFD support structure with the staggered configuration exhibited the highest yield (∼55%) for light olefins at 350 °C and high WHSV. This was due to improved mass and heat transfer properties.

Van Noyen et al.¹¹⁹  3D-printed porous titanium grade 5 (Ti6Al4V) alloy using 3DFD technique (DIW). In this example, two types of pore architectures: straight (1–1) and staggered (1–3–5) were prepared and subsequently coated with ZSM-5 by immersing support structures in the synthesis solution in an autoclave and performing zeolite synthesis. These coated support structures were subsequently tested for the decomposition of nitrous oxide. While both types of structures were highly active and stable, staggered (1–3–5) stacking gave slightly higher conversions. Additionally, when pressure drop was measured and compared against a packed bed of beads (3.2 mm), it showed that both 3D-printed structures had a lower pressure drop than the packed bed. The least pressure drop was observed for straight (1–1) stacked structures which can be expected where the path is not as tortuous as staggered (1–3–5). This study showed that using a 3DFD technique coupled with a zeolite catalyst can allow obtaining a structure with hierarchical porosity. This porosity can further be tuned from macro to micro or meso to micro scale. The technique being generic has wide applicability to various materials such as ceramic or metallic and to various reactions.

Chaparro-Garnica et al.¹³⁰  modified VisiJet FTX Green polymeric resin (3D Systems Inc.) with different amounts of carbon (0.13 wt%) and silica (0.35 wt%) and printed it in honeycomb monolith shapes. In addition to the conventional flat channeled monoliths, the authors also printed a monolith with prismatic cavities (advanced design) to improve the anchoring of the powdered catalyst. On these two monolith designs, CuO/CeO₂ active phase was deposited with dip coating and tested for preferential CO oxidation reaction. The advanced design and doping with carbon and silica improved the anchoring of the active phase in a single impregnation. Although the monolith-supported CuO/CeO₂ showed lower activity than powdered CuO/CeO₂, the activity was enhanced after multiple reuse cycles paving the way for potential application in PEMFC (Proton Exchange Membrane Fuel Cell).

In another work, Chaparro-Garnica et al.¹¹⁵  used an extrusion-based method to 3D print polymeric templates that were further used to create integral carbon monoliths in two designs: conventional straight channeled and advanced design-based with tortuous paths. The active phase of Ni/CeO₂ was applied on these carbon monoliths by dip-coating and the monoliths were further tested for CO₂ hydrogenation into methane. Advanced design-based monolith catalysts provided high conversions (80%) as compared to 60% for conventional straight-channeled structures at 300 °C. These monolithic catalysts were also found to be highly stable, reusable, and reached equilibrium conversions at approximately 350–400 °C.

Konarova et al.¹¹⁴  3D-printed carbon supported NiMo-based catalysts using an extrusion-based method and compared against activated carbon (pellet) supported NiMo catalysts. These NiMo catalysts were 3D-printed in two orientations: square patterned and honeycomb patterned using PVA and starch as a carrier, and further tested for CO hydrogenation reactions. These catalysts showed higher CO conversions and lower byproduct selectivity (for methane) as compared to conventional pelletized catalysts. Additionally, the drop in the activity was significantly more for pelletized catalysts when the gas hourly velocity was increased. This was because of favorable diffusion paths. Depending on the polymer used as the carbon source, various microstructural features could be obtained.

Zhou and Liu¹³¹  3D-printed porous carbon structures with tailorable pore sizes. They used starch/gelatin ink as a carbon source and SiO₂ monodispersed sphere as a hard template. This allowed them to get accurate structures with meso/macro porosity. This porous carbon structure was further used for the oxidation of benzyl alcohol. Results indicated that the 3D modeling controlled the macro-scale structure whereas the SiO₂ template controlled the macro-meso pore structure. Additionally, SiO₂ particles were found to help rheological properties and also in reducing shrinkage during carbonization. 3D monoliths with large porosities showed high conversions (90%) whereas the monoliths with no open porosity showed low conversions (70%) at high benzaldehyde selectivity (∼90%).

A catalytic reactor is a device where the chemical reaction(s) takes place, with the catalyst incorporated into the system. As most of the catalysts are in solid form, in the industry of heterogeneous catalysis, fixed-bed and fluidized-bed are the most used gas–solid two-phase catalytic reactors. For liquid–solid or gas–liquid–solid catalytic reactors, fixed-bed (including trickle-bed) and slurry reactors are used the most. Despite the maturity of reactor assembling technologies, reactors are still voluminous, centralized, and inflexible. There is an increasing demand for customized reactor design from laboratories and industries. However, current technologies do not meet such demands well due to (1) high cost of customized reactors, (2) the long wait time for the reactor production process, and (3) the limited availability of the reactor manufacturing equipment in most of the laboratories and chemical industries. These limitations are more significant in the development of flow reactors and microreactors. That is to say, the development of a reactor for specific catalytic reactions is not widely accessible, and that is one of the driving forces of the development of advanced reactor manufacturing technologies which aim to provide a cheap, flexible, and feasible way to fabricate new and unique reactors for specific needs.

The development of 3D printing and AM technology has recently been applied in the field of catalysis, including catalyst synthesis and reactors. Compared with conventional reactor manufacturing, 3D printing facilitates the production of complicated reactors without the use of a mold or template and also provides high reproducibility due to the minimal manual reactor assembly involved.^132–134  Since 3D-printed catalysts have been discussed earlier in the previous section, this section will focus on different 3D-printed catalytic reactors including “self-catalytic” reactors.

The general process for fabricating a 3D-printed reactor is to design the reactor structure through a computer-aided design (CAD) process. The design file will then be delivered to the AM machine, and the target material needs to be prepared. Sometimes a post-processing step, such as support material removal, is necessary.¹³⁵  There are several different 3D printing techniques including direct ink writing, fused deposition modeling, electron beam melting, etc., and the selection of the 3D printing technologies is based on the requirement of the reactor material. The interested reader is directed to these review articles for a more detailed discussion on these technologies.^5,136  Typically, the material of 3D-printed reactors is polymers (including thermoplastics), ceramics, metals (including stainless steel, aluminum, etc.), and photoresins (including epoxy-based resin and acrylate). In general, 3D printing does not show an advantage for batch reactors because the conventional molding process and the designs remain simple and consistent.

3D printing technology enables various reactors to be “home-made”, such as reactors with customized shapes, those run under special reaction environments, and reactors with sophisticated details.^137,138  A catalytic static mixer (CSM) reactor can act as a flow reactor with enhanced surface area for catalyst doping as well as improved mass and heat transfer due to the special geometries of the static mixer insertion.¹³⁹  The catalyst can be deposited on the mixer insertion via various methods such as cold-spraying. The material for CSM reactor is usually stainless and can withstand up to 250 °C, as claimed by Hornung et al.¹⁴⁰  Fig. 3 shows an example of CSM with catalyst deposition. Compared with the conventional fixed-bed reactor, the fluid flow in the CSM reactor is more regular and the pressure drop is lower.¹⁴¹  Zhu et al.¹⁴²  compared the catalytic performance of vinyl acetate hydrogenation, and found that applying a CSM reactor can overcome mass transfer limitation which occurs at a high hydrogen-to-substrate ratio. Also, the fluid flow in the CSM reactor is more predictable by using computer-aided simulation compared with a conventional fixed-bed reactor. This setup has been widely applied in organic processes (e.g., hydrogenation, transesterification).^139–146 

Microreactor fabrication is a major field in which 3D printing can play an important role. Microreactors consist of narrow channels that provide a large surface area-to-volume ratio, and this feature can bring several advantages to catalytic reactions, such as enhanced heat and mass transfer, better control of reaction conditions, and prolonged contact between fluid and catalyst.¹⁴⁷  The concept of a catalytic microchannel reactor has been developed over decades, and there are many research projects that apply this design to several catalytic reactions such as water-gas shift,¹⁴⁸  biofuel production,¹⁴⁷  and various organic synthesis processes.¹⁴⁹  Due to the limitations of commercial microfabrication techniques, the price of microreactors is generally high.¹⁴⁹  3D printing enables a “homemade” microreactor which has the potential to reduce the cost and complexity of setting up a microreactor system. Kuila’s group developed a 3D-printed stainless steel microreactor for the Fischer–Tropsch (FT) synthesis with different catalysts coated on the wall of microchannels (Fig. 4).^150–153  They found that compared with silicon microreactors manufactured by microfabrication technique, a 3D-printed stainless steel microreactor not only has a higher mechanical strength enabling high-pressure operations for future process development but also improves heat transfer which facilitates the control of highly exothermic FT process.¹⁵³  Kucherov et al.¹⁵⁴  fabricated metal-coated microreactors with different channel profiles and applied them in both heterogeneous and homogeneous (catalytic) organic synthesis processes. They found that such designs can sustain aggressive reaction conditions, eliminate imperfections of 3D printing, and improve the mechanical properties of the reactors.¹⁵⁴  A microreactor can also be in a porous tubular form with complex internal geometries, and the 3D printing technique can easily fabricate such designs. Baena-Moreno et al.¹⁵⁵  compared the catalytic methane dry reforming performance over a conventional honeycomb monolith reactor and 3D-printed complex gyroid monolith. The result showed that the 3D-printed gyroid monolith outperformed owing to enhanced mass and heat diffusion processes.¹⁵⁵  Similar enhancement in catalytic performance and transport phenomena can also be found in liquid-phase catalytic reactions using 3D-printed porous tubular microreactors.¹⁵⁶ 

3D-printed self-catalytic reactor (SCR) is another unique flow reactor that has attracted much interest recently. Unlike the reactor systems mentioned earlier, SCR combines reactor printing and catalyst preparation minimizing the cost while improving energy efficiency. This reactor design is generally better suited for gas–solid heterogeneous catalytic reactions such as FT synthesis and methane dry reforming.^157,158  For instance, according to Wei et al.,¹⁵⁸  the 3D-printed honeycomb-like Fe-alloy tube serves as both a catalyst and a flow reactor for FT synthesis to produce liquid fuels. Printing different internal profiles within the reactor varies flow properties enabling facile control on FT reaction product distribution.¹⁵⁸  The research on similar self-catalytic systems is still novel and limited.

Since Fujishima and Honda first discovered the photocatalytic effect of TiO₂ on water decomposition, photocatalytic reactions have stepped into the spotlight due to the sustainability of the process.¹⁵⁹  In the recent decade, photocatalysis has been applied in pollution control,^160,161  contaminant mitigation,^162,163  methane conversion,¹⁶⁴  hydrogen production,^165,166  carbon dioxide conversion,^167–171  etc. The most widely-applied catalyst support in photocatalytic reactions is TiO₂, and the reaction mechanism (for pollutant removal) is the production of hydroxyl and superoxide radicals from the photoelectronic effect between light irradiation and optical semiconductor catalyst support, and this mechanism is a well-established.¹⁷² 

For a liquid–solid heterogeneous photocatalytic system, there are two major reactor configurations based on the relative motion of the photocatalyst: immobilized and slurry. Immobilized reactors have a photocatalyst fixed on a support or dispersed onto the stationary phase which enables continuous operation.^173,174  The slurry configuration is when the photocatalyst particles are dispersed or suspended in the fluid in the reactor to enhance the total surface area of the photocatalyst per unit volume. Compared with continuous immobilized reactors, slurry reactors generally have a more uniform catalyst distribution and higher catalyst-fluid interaction surface area at the same reactor volume settings, which results in better performance, but an extra after-process separation step is required.^174,175  Conventional photocatalytic reactor designs have an issue of low efficiency due to limited mass transfer, so customized reactor designs, including using microreactors, can be applied to solve this problem.

3D printing technology can be therefore integrated into such customized photocatalytic reactor designs. This application is new and there are only a few pioneering research works demonstrating the advantages of using a 3D-printed liquid–solid photocatalytic reactor. Zhou et al.¹³⁸  3D-printed two lab-scale sinusoidal plastic photocatalytic reactors with different settings, and successfully removed pollutants (i.e. methylene blue and phenol) from water. Such design protocol has the advantages of high flexibility, quick fabrication times, and cost-effectiveness. Phang et al.¹⁷⁶  used different resins to manufacture a reactor chamber for photocatalytic wastewater treatment. By using high heat deflection temperature resin as the chamber body and Eco UV clear resin as the cover, the 3D-printed photocatalytic reactor can withstand high temperatures (up to 300 °C) and adequate light penetration (92% for LED light).¹⁷⁶  By using g-C₃N₄ immobilized catalytic coating, the result shows that the integrated 3D-printed reactor improves cost and time-effectiveness by eliminating the number of unit operations (process intensification see Section 5).¹⁷⁶  Pellejero et al.¹⁷⁷  prepared transparent microchannel reactors for continuous photoreduction of 4-nitrophenol with NaBH₄ (Fig. 5), and the catalyst can be easily deposited in the channel by simple dip-coating method. Such a design extended the residence time and improved photocatalyst distribution, which resulted in high catalytic performance.¹⁷⁷  Other than the reactor itself, 3D printing can be applied to manufacture the component of the reactor system. In larger-scale photocatalytic reactors, parabolic concentrators are required for tubular reactors to enhance the light intensity of the systems. However, the manufacturing of the concentrator requires precise engineering and special processes.¹⁷⁸  In this case, using 3D printing to manufacture the concentrator can lower the cost of setting up a conventional tubular photoreactor, while optimizing the geometry simultaneously.¹⁷⁸ 

Intensification of processes largely relies on factors such as structure or geometry, size, and the type of material used to make the catalytic devices, which are all addressed in the design. As a result, if process intensification provides a link between catalysis and manufacturing, design is the instrument that consolidates this relationship. It begins with the demands of catalytic processes and allows for the consolidation of ideas through AM.

AM is a novel manufacturing technique that allows the flexible creation of intricate and precise 3-D geometries that would be impossible to accomplish using standard fabrication processes such as casting and machining.¹³⁵  The unique ability of AM technologies to transform materials into functional devices with a specific geometry has sparked interest in a wide variety of fields to provide custom-made designs for tailored applications such as product development, manufacturing aids, as well as end parts, among others.⁶ 

Process intensification (PI) is considered to be one of the most promising progress paths for the development of more sustainable chemical processes. From the standpoint of chemical processes, the goal is to create renewable fuels and value-added products while maintaining optimum productivity with minimal energy consumption, economic investment, and CO₂ footprint.¹⁷⁹  PI is defined as a set of innovative principles applied in process and equipment design, which can bring significant benefits in terms of process and chain efficiency, lower capital and operating expenses, higher quality of products, lower wastes, and improved process safety.

Among the most prominent illustrations of PI is the employment of microchannel reactors to carry out procedures on both a laboratory and an industrial scale.^180,181  Because of the submillimeter size of their channels, microchannel reactors boost both heat and mass transport phenomena, allowing reactions to occur quicker, more selectively, and with better energy efficiency in miniaturized volumes. However, as recently emphasized by Konarova group,¹¹⁴  the broad implementation of microchannel reactor technology is impeded by the reactor’s complicated design, high manufacturing cost, and low volume ratio of catalyst and the reactor. However, the same authors point out that AM might remove such barriers.

Parra-Cabrera et al.¹⁰  proposed a relationship between catalysis and AM. They stated that the most promising prospects for merging process intensification with AM are in continuous flow processes that employ heterogeneous catalysts and/or have transport constraints. The advancements in this subject would thereafter be determined by increasingly complicated situations of non-isothermal reactions and multi-phase settings.¹⁰  Hurt et al.⁶  proposed that there is an incentive to implement AM in heterogeneous catalysis (although it could well be extrapolated to catalysis in general) based on two fundamental motivations: (i) to scale down the reactors as a sustainable approach to producing chemicals on demand on-site. This is based on the fact that miniaturization does not have to imply a decrease in conversion, as demonstrated by Tubio et al.¹²⁶  with their printed monoliths with Cu-based catalyst tested in various Ullmann reactions; and (ii) secondly, large scale chemical manufacturing units must be adapted to a mode of operation that minimizes environmental impact while maximizing energy efficiency.

Even though this is a challenging and expensive effort, the considerable adaptability that AM provides may allow huge enterprises to achieve the progressive transformation they desire without having to design and build their facilities from scratch. As may be deduced, these two reasons are completely consistent with process intensification concepts. As a result, the combination of AM with catalysis via process intensification may provide several benefits to the chemical sector.

AM technology has a unique advantage in its ability to address the integration of catalysts into structured materials. The technologies can produce complex 3D structures of smaller dimensions and, in some cases, high surface roughness via novel techniques such as 3D printing (3DP).

The first type of material is polymeric. According to Zhou et al.,¹³⁶  this is the most extensively utilized group in AM since it can be employed in extrusion, light polymerization, power bed, and practically all other popular printing processes. These materials have limited surface areas and poor surface characteristics, making them difficult to incorporate into catalytic solutions regardless of the method. One approach to overcoming such disadvantages is to include active chemicals in the polymeric matrix to develop hybrid materials such as TiO₂-ABS, which would expand the use of doped polymers in the printing of catalysis systems. Nonetheless, additives are essential to maintain the appropriate rheology for an accurate printing process, and managing the effects of such additives is challenging, particularly given the nature of the catalyst if those additives are present during the printing process.⁷ 

Given the facts, printing catalysts or catalytic supports using polymeric materials does not appear to be beneficial. However, there are alternative ways to apply AM of polymers to the field of catalysis, and it is conceivable to facilitate the catalytic process by printing reactors with improved designs. As a result, a series of situations in which AM has been used to create appropriate reaction ware for catalytic processes are presented and will be discussed below focusing especially on printed reactors for catalytic reactions.

Another method for incorporating AM into the production of ceramic structures is to employ polymeric printed molds with sophisticated designs. The ceramic material is blended with additives to create a paste that is implanted within the template, then calcination is used to gasify the mold material and solidify the ceramic piece. This technique is known as indirect printing. This has been used to create structured systems for catalytic purposes. In the case of Davo-Quinonero et al.,¹⁸²  they printed templates of monoliths with a commercial ultraviolet curable polymeric resin (Visijet FTX green), which were then filled by extrusion with a commercial cordierite paste (COR-MIK-MP).

Carbon compounds have intrinsic characteristics such as large surface area, allotropy, electrical conductivity, and chemical stability, making them appealing as catalytic supports and hence printing materials.¹³⁶  However, because of their properties, they must be treated similarly to ceramic materials since they require additives to form a matrix with suitable rheology for the printing process.

Carbonization of thermosetting resins is another option for generating carbon-printed products since these materials have a high degree of crosslinking, which can prevent print distortion during the carbonization process. Resorcinol–formaldehyde (R–F) solutions are an excellent example of a pre-polymer solution with the appropriate rheological behavior (non-Newtonian fluid with high viscosity) that permits a regulated chain growth process in an alkaline medium.^136,183  Following a DIW process, these pre-polymer solutions can be directly injected through a nozzle as seen in Fig. 6, or they can be incorporated in a printed template that dictates the geometry of the final structure after a carbonization process.

Because so few investigations in this sector, bioactive materials (bio-printing) aims to incorporate the benefits of process intensification by developing systems. In this view, immobilizing bioactive species in a structured system would enable their extraction from the reaction media, which is one of the most difficult hurdles in enzymatic processes.⁷  Nonetheless, bio-active printing necessitates significantly more precise production and post-treatment conditions than those detailed thus far to maintain the bio-active agent alive and completely active throughout the process.

For immobilizing catalysts in a 3D-printed structure, two ways are used: integration and functionalization. The catalyst is included in the print medium by either mixing it in as an additive or printing the catalytic material directly.¹⁰  Although polymers are the most often used filament printing medium, pure polymers are seldom active catalysts. Instead, additives such as metal oxides, nanoparticles, or metal–organic frameworks (MOFs) can be printed.¹³⁶  Post-print processing and chemical modification are the second methods for immobilizing catalysts.^122,184 

For process intensification, engineers may design structures out of inert or useful materials because of the range of the AM materials library and manufacturing procedures. Printing the catalyst material directly is a way to decrease pre- or post-processing in heterogeneous catalysis. A recent study used selective laser sintering (SLS) to create self-catalytic reactors (SCR) out of iron, cobalt, and nickel,¹⁵⁸  where metal 3D printing products can simultaneously serve as chemical reactors and catalysts for direct conversion of C₁ molecules (including CO, CO_2, and CH₄) into high value-added chemicals. For example, the Fe-SCR and Co-SCR successfully catalyze the synthesis of liquid fuel from Fischer–Tropsch synthesis and CO₂ hydrogenation; the Ni-SCR efficiently produces syngas (CO/H₂) by CO₂ reforming of CH₄. The influence of the Co-interior SCR’s geometry was explored and it was discovered that the structure had a significant impact on catalytic selectivity.¹⁵⁸  Other approaches to printing metal and metal oxide catalysts have been studied. Zhu et al.¹⁸⁵  printed AgAu composite inks to construct nanoporous structures that increased methanol oxidation mass transfer. Rapid screening of mixed metal oxide compounds for fuel and electrochemical applications has been accomplished using inkjet printing.^186,187  Because of their catalytic activity and capacity to promote catalysis, graphene, and its derivatives are materials of great interest in the catalysis field.¹⁸⁸  Zhu et al.¹⁸⁹  used 3D printing to create periodic graphene aerogel micro-lattices. The developed materials have enormous surface areas, low density, and mechanical properties. The composition of the printed ink might be used to tailor the qualities. Photopolymerization stereolithography was utilized by Manzano et al. (SLA) using catalytically active functional acrylate monomers.¹⁹⁰ 

Stucker et al.¹³⁵  established the effect of catalyst shape on throughput and conversion. They developed two ceramic catalytic fixed beds, one extruded in a honeycomb configuration and one robust in a lattice geometry. In comparison to the honeycomb construction with straight-through channels, the complicated internal geometry of the lattice induced turbulence, increased mass transfer to the catalyst surface. The lattice fixed bed converted six times more methane during catalytic combustion. Additive printing can be utilized to control the porosity of the catalytic monolith and shift the selectivity in the desired direction in reactive systems with competing surface and bulk phase reactions.¹⁹¹  Catalytically active mixing devices for both flow¹⁴¹  and batch reactors¹⁹²  have also been created via 3D printing. This method is getting increasingly popular among AM researchers for its enhanced turnover efficiency as well as the ease of use in process.¹⁹³ 

Multiphase chemical systems are an essential component of the catalyst industry, and more emphasis and effort in advanced manufacturing methods are required to reduce the cost of catalyst production, reactor components, and energy input while also lowering the carbon footprint. Although some of the research into the application of these methods for catalytic processes is still in its early stages, promising results have been obtained. These methods allow researchers to synthesize the desired material for running the processes in optimized conditions, and also regulate catalyst and device geometries at many length scales. This allows the construction of novel catalytic systems that maximize process rates and provide a uniform molecular experience. Therefore, we can expect that advanced manufacturing methods will continue to develop as a promising platform to intensify catalytic reaction processes and enable a range of sustainable production operations. Despite immense progress, there still exist some challenges that need to be addressed in the coming years as described below.

1. Lack of high precision and large-scale control over advanced coating processes is one of the challenges.

2. Lack of automation in catalyst production that can increase the line speed, and minimize operation cost and energy utilization in catalyst production.

3. Lack of a significant number of AI/ML applications to catalytic studies although it is on the rise.

4. Lack of flexibility in printability of AM materials (not all are printable on every AM platform), the variation in resolution/tolerance of 3D-printed structures, the higher number of post-processing operations, anisotropic properties of printed material, thermodynamic stability, and inability to print multi-material components.

5. For additively manufactured reactors, a limited number of scale-up and reliability studies, and a lack of proven longevity tests are some of the challenges.

This project was funded by the U.S. Department of Energy, National Energy Technology Laboratory, in part, through a site support contract. Neither the United States Government nor any agency thereof, nor any of their employees, nor the support contractor, nor any of their employees, makes any warranty, express 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. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.