A large fixed bed reactor for MRI operando experiments at elevated temperature and pressure

Process intensification of chemical reactions is one of the main tasks in chemical engineering aiming to increase the efficiency of production steps.¹ In this context, many chemical reactions are optimized toward higher efficiencies driven by economic and ecologic aspects.² Currently, the optimization of power-to-gas or power-to-liquid technologies (PtX) is widely discussed in the literature due to their potential to store excess renewable energy.^3–5 A prominent example is the CO₂ methanation reaction (or the Sabatier process), where carbon dioxide reacts with hydrogen to give methane and water as follows:⁶ 

This gas-phase reaction takes place on the solid surface of a catalyst⁷ while releasing heat. Conventionally, this can be done in tubular fixed-bed reactors where the active catalyst material (e.g., Ru or Ni) is impregnated on porous pellets. Alternatively, the catalyst can also be applied on monolithic structures such as honeycombs or open-cell foams.⁸ In order to boost the productivity of the overall reaction system (i.e., reactor and reacting conditions), it is crucial to understand macroscopic transport processes of heat, mass, and momentum.⁹ For example, the macroscopic heat transport influences the interplay between kinetics and thermodynamics, which influence the yield and selectivity of the chemical reaction.¹⁰ Spatially resolved profiles of temperature, gas composition, and velocity are crucial for detailed catalyst studies. Relating temperature and gas composition to the local catalytic behavior will benefit the understanding of the reaction and reactor development.

State-of-the-art measurement techniques, however, are often not able to resolve local profiles, while an optical measurement is usually hindered by the opaque catalyst support material. One dimensional spatial resolution can be achieved by inserting a capillary probe, such as the SpaciMS system,^11,12 infrared spectroscopy,¹³ or electrical resistance tomography.¹⁴ For catalytic insight, two dimensional infrared thermography can be used, which requires an optically accessible window.^15,16 So far, only the application of Nuclear Magnetic Resonance (NMR) techniques has shown the ability to measure operando 3D spatially resolved profiles. While liquid-phase NMR is more established, gas-phase NMR suffers from a lower number of hydrogen nuclei (i.e., spin density) and lower signal intensities and is, thus, considered more challenging. Recent studies mapped temperature, concentration, and velocity distributions inside porous media in the gas phase^17–23 but are yet limited to model reactions taking place at industrially irrelevant conditions, i.e., ambient pressure and temperature (e.g., ethene hydrogenation). In contrast, industrially relevant heterogeneous gas-phase reactions usually require elevated temperatures and pressure ranges (e.g., for the CO₂ methanation, p = 4–10 bars and T > 250 °C).²⁴ These requirements contrast one fundamental constraint of NMR studies: the region of interest (ROI) must be free of electrically conducting bulk material. Furthermore, NMR devices are quite sensitive to increased temperatures due to their numerous electrical components. This culminates in the requirement of an NMR-customized reactor setup, which ensures reactor security at elevated pressures and temperatures without using electrically conducting bulk metal in the vicinity of the ROI. Recently, the group of Gladden²⁵ reported the usage of such a device, but it is neither commercially available nor a detailed description published.

Overcoming the design challenges for NMR is imperative in order to advance in the rapidly growing field of operando NMR measurements. We, therefore, present an open concept of a chemical reactor for NMR experiments suitable for a wide array of applications. We share the decision criteria, which enable the reactor to endure temperatures above 400 °C and pressures up to 30 bars. We further provide sketches, suppliers, and Autodesk Inventor 2018 CAD files of each part in the repository.²⁶ Furthermore, we will share tests of the reactor performance regarding pressure tightness, magnetic shielding, and temperature stability. Its applicability is demonstrated by performing the CO₂ methanation in the developed setup.

A common reactor type in heterogeneous catalysis from lab to pilot scale is the tubular reactor (Fig. 1), with a catalyst bed in which the reaction takes place. To remove the evolving heat during an exothermal reaction, a concentric shell around the tube is often added, allowing the use of a coolant. The design of the reactor concept presented here is based on this type and has a similar size. As already mentioned, NMR tomographs and the NMR technique pose certain restrictions on the investigated materials and setups. Guidelines on building magnetic resonance imaging (MRI) compatible setups are scarce and always subject to their specific applications. In our work, we have identified the following constraints:

• Free of magnetic parts: Due to the strong magnets required in NMR tomographs, the setup and periphery must be completely free of any magnetic materials (i.e., no magnetic material in the experimentation room).

• Region of interest (ROI) free of metallic materials: Due to the fast switching electric fields required to obtain 3D MRI data, the ROI must be free of electrically conducting material.

• Withstand elevated pressures and temperatures: Depending on the investigated reaction, the setup must be able to withstand elevated pressures and potentially high temperatures in the reactor bed.

• Outside temperature of the reactor of less than 80 °C: As the coil that surrounds the outer reactor wall is sensitive to elevated temperatures, the temperature at the outermost shell of the reactor should not exceed 80 °C. The coil used in this work is intended for high temperature usage, and other coils might only withstand an inner temperature below 50 °C.

• Outer diameter at most 111 mm: Depending on the NMR magnet used, this value might be different. The coil and reactor must fit into the bore of the NMR tomograph, which limits possible space for the isolation material and a potentially larger reaction zone.

Material and design choices have to be made to accommodate these constraints. This chapter gives a brief overview on the tradeoffs between different ranges of reactor performances and the corresponding decision criteria, which eventually led to the proposed reactor design (Fig. 2).

Tubular reactors are commonly made of stainless steel. It is indeed possible to perform NMR measurements inside metallic vessels, and a variety of setups built for higher pressure can be found in the literature, often involving titanium as a key component.^27–29 However, to perform 3D measurements of the reaction zone using MRI, it is necessary to rapidly switch magnetic field gradients from the outside. This requires the region of interest (ROI) to be free of electrically conducting bulk material and, thus, free of metal parts. For this reason, glass is commonly used for MRI setups. Although glass is a suitable material for MRI studies,³⁰ it requires comparably large wall thicknesses in order to withstand elevated pressure. Another application reports an MRI vessel made from polyether ether ketone (PEEK) and brass that is used at 200 bars.³¹ However, the mechanical stability of PEEK significantly decreases above its glass transition temperature. Furthermore, PEEK contains hydrogen atoms, which might overshadow the comparatively low gas phase signal.

In this study, we use a ceramic tube for the main reaction zone in the center (inner tube, Fig. 2). The decision was made because most ceramic materials show non-metallic behavior in terms of electric conductivity. Furthermore, ceramics normally have high stress resistance, while suffering from low fracture toughness. While assembling the ceramic reactor and catalytic fixed bed, torsion might cause heavy stress for the material. To reduce the possible tension on the tube during assembly and application, the ceramic tube is kept between two flanges (inner flange, Fig. 2) only being held by one O-ring at each end. The flanges are made of titanium due to its excellent mechanical properties compared to other non-ferromagnetic metals, such as aluminum. The length of the tube ensures the ROI being sufficiently far from the flanges such that the titanium does not influence the measurements.

A fastener kept the flanges in place and retained the pressure in the axial direction. It consisted of secondary flanges (outer flange, Fig. 2) and custom-made fiberglass [glass reinforced plastic (GRP)] rods. To be able to fasten the rods, aluminum sleeves with threads are glued (3M Scotch-Weld DP 490, 3M Deutschland GmbH, Neuss, Germany) onto the GRP rods. The secondary flange is made of polyamid imide as its low thermal conductivity in comparison with that of titanium and other metals reduces heat transported in the radial direction toward the NMR magnet and the radio frequency (RF) coil. Polyamid imide exceeds the mechanical properties of many other polymers and has one of the highest glass transition temperatures, thus making it exceptionally fit for high temperature and high pressure applications. To minimize the transport of evolving heat at the reaction zone toward the outer parts, the whole tube is covered in a thick glass fiber mat (isolation, Fig. 2). A glass tube around the isolation material adds mechanical stability and reduces heat conduction even further.

Finally, a special MRI coil was built to fit the setup. The coil was developed and manufactured by MRI.Tools GmbH (Berlin, Germany) using a ring-shaped cylindrical fiberglass body with four cylindrical bores distributed evenly on the ring, making room for the GRP rods. The rods thus go through the coil’s body. Furthermore, connections for pressurized air were added at the front to cool the coil in situ through thermal convection. Compared with previously used standard RF coils, the dedicated RF coil built by MRI.Tools exhibits reduced background signals and considerably reduced detuning of the RF coil with changing temperature, both properties being essential for spectroscopic MRI (MRSI) measurements of gas-phase reactions.

The final reactor setup is shown in Fig. 3. The flanges and the GRP rods have been manufactured in university facilities. All other parts were bought directly from external suppliers ( Appendix A). The total costs for the materials used are about 3000 € (without coil), when applying current market prices (2019). The original design involved a shell around the ceramic tube for cooling, but to reduce the complexity of the first tests, it was chosen to use the version described above.

To ensure that the setup is working as intended, we performed a pressure test and an analysis of magnetic shielding. The applicability of the reactor was then demonstrated for the CO₂ methanation, a strongly exothermic gas-phase reaction.

First, the stability of the gaskets and the inner tube was tested for pressures up to 28 bars. To this end, we increased the pressure to 28 bars and held it over an hour. Prior to the test, the reactor was filled with water to reduce the effective gas volume, which limits the potential damage in the case of failure. Subsequently, we increased the pressure to the targeted value of 28 bars and closed the valves before tracking the temporal evolution of the pressure for an hour. For a better readability, the pressure drop was calculated into a leakage rate, as shown in Fig. 4. When scaled to the actual reactor volume, the total pressure loss over 1 h was 0.05 bar (<1%), ensuring satisfactory pressure tightness of the proposed reactor design and safe usage up to the mid pressure range of 10 bars. Tests at higher temperatures and combined temperature and pressure have not been carried out yet.

In most NMR applications, specimens are presented either directly to the RF coil or inside the so-called NMR tubes that are thin glass tubes for liquid storing. To deal with the increased pressure and temperature, the amount of solid material inside the MRI coil needed to be significantly increased compared to the thin-walled NMR tubes. Even though the inner tube, glass tube, and isolation material are not magnetic, we expect that the solid material influences the magnetic field inside the coil. This may decrease the effective decay time of the signal response (⁠$T2*$⁠) and, therefore, reduce the ability to distinguish different NMR spectra. This is because the line width [full width at half maximum (FWHM)] is given by $1/(πT2*)$⁠.

To test this, a 3D spectroscopic imaging (3D-MRSI, see  Appendix B) sequence was used to measure pure methane inside the inner tube of the reactor and inside a gas sampling container made of glass. The gas sampling container was filled by flushing the container for over a minute with pure methane and then sealed by closing both valves simultaneously. The same procedure was performed on the inner tube by flushing the tube for over a minute. Since the inner tube cannot be sealed without applying pressure, a small volume flow of 0.5 L_N min⁻¹ methane was flushed through the tube (superficial velocity = 0.0257 m s⁻¹), which is equivalent to non-flowing methane for this type of measurement. The measurement of the gas sampling container resulted in a $T2*$ value of 26.16 ms, which is equal to the longitudinal relaxation time (T₁) of methane. At ambient temperature and pressure, the T₁ value of methane is equal to the T₂ value,³² and thus, $T2*$ is also equal to T₂, proving the absence of magnetic shielding effects of the glass sampling container. Measurements of methane in the inner tube of the reactor instead resulted in a reduced $T2*$ value of 21.3 ms. The obtained spectra show a slightly increased peak width inside the reactor (FWHM = 14.94 Hz as compared to 12.17 Hz; Fig. 5), but the difference is almost invisible, and thus, no major material influence is assumed.

To demonstrate the applicability of the developed reactor under operating conditions, we performed the CO₂ methanation reaction [Eq. (1)] using a Ru/Al₂O₃-coated honeycomb catalyst. A description of the catalyst synthesis procedure can be found in  Appendix B. Considering that electrical heating inside the NMR system is not possible, we instead preheated the reactor by using the hydrogenation of ethene³⁰ to achieve the required onset temperature of the methanation (T_start ≈ 250 °C). The hydrogenation of ethene is almost as exothermic as the methanation reaction but can be initiated at ambient temperature on the same catalyst. The reactor was preheated using a 5 L_N min⁻¹ gas flow (H₂:C₂H₄, 1:1) for 1 h. Once the required temperature was achieved, ethene was substituted by CO₂, and a 5 L_N min⁻¹ gas flow (H₂:CO₂, 3:1) was set. The methanation reaction was then run autothermally for almost 2 h to ensure reaction stability. To visualize methane and to show 3D characterization of gases inside the reactor, we performed operando 3D-MRSI (i.e., spectroscopic imaging) measurements. Additionally, we tracked the formed methane by measuring the output flow with a non-dispersive infrared spectrometer (NDIR spectrometer, GMS810 with a MULTOR module, Sick AG, Germany). After switching from the hydrogenation of ethene to CO₂ methanation, the output volume concentration of methane sharply rose to about 18 vol. % and over time gradually increased to a maximum of about 30 vol. % (Fig. 6). During this period, a series of 10 3D-MRSI measurements were carried out, each with a duration of about 8 min.

It was possible to process the obtained MRSI spectra into 3D fields of methane (Fig. 7), which resemble very well the honeycomb structure and show the successive increase in methane formation both spatially—in the flow direction—and temporally resolved over each measurement (Fig. 8).

The methane signal amplitude depends on both the concentration and temperature. Despite the temperature dependency, the mean signal amplitude at the last slice of each 3D-MRSI measurement corresponds very well to the concentration measured using the NDIR spectrometer. Against our expectations, the signal amplitude is not constant in the y direction in the first half of the 3D field of methane. As the honeycomb is a symmetrical structure, the possible influences on the signal amplitude from local temperature differences should be symmetric as well. Instead, this effect might be a result of fluctuations due to the low SNR of gases or inhomogeneous catalyst loading. While a detailed analysis of the underlying effects is beyond the scope for this study, the occurrence highlights the possible gain of information through operando spatially resolved NMR studies. We would further like to note that the reactor protected the surroundings very well from the extreme internal conditions. For safety reasons, the temperature at the inner bore of the MRI coil, only a few centimeters outside, was measured using a fiber optic temperature sensor (Weidmann Electrical Technology AG, Rapperswil, Switzerland). In an autothermal state, temperatures inside the honeycomb can rise to 400 °C or more during exothermal reactions,¹⁰ but the temperature outside of the reactor was safely kept below 65 °C. For further details of the reaction setup of the CO₂ methanation, see the corresponding section in  Appendix B.

In this work, we present the design and testing of a reactor that enables MRI measurements of chemical reactions at elevated temperature and pressure. The concept of the reactor is simple enough to be adapted to different demands. By changing materials or tube diameters, the pressure stability can be fitted to a desired reaction. Likewise, it is possible to add a shell for a temperature control fluid to cool the reactor, as commonly used in lab or pilot scale reactors. Using this concept, users should be able to build a setup of their own, as all parts either are commercially available or can be manufactured in conventional workshops. The polyamide imide made up for about one third of the price. With less challenging constraints than necessary for the CO₂ methanation concerning either temperature or pressure, the secondary flanges can be made of different materials and the setup cost can be reduced significantly.

The tests on the developed reactor demonstrated that this design can serve as a key component for MRI on chemical reactors. In the proposed configuration, the reactor withstands a pressure of at least 28 bars and showed little leakage during the test (Fig. 4). The reactor withstood the high temperature of the autothermal methanation reaction without any signs of fatigue due to the increased stress and with low thermal conductivity toward the outside. However, pressure tests at high temperatures have not been carried out yet due to safety concerns in the case of breaking. Due to the low influence of the reactor on magnetic field homogeneity, we could demonstrate the applicability by performing operando 3D-MRSI measurements of the CO₂ methanation reaction.

This reactor is a step toward spatially resolving the temperature, concentration, and velocity distribution during chemical reactions at operating conditions (i.e., elevated pressure and temperature). With this at hand, we are now able to study the physical phenomena inside catalyst beds directly and without external influences. By sharing this work, we hope to encourage more groups to start their research on MRI of chemical reactions and gas-phase reactions, in particular, and to further reveal the secrets behind chemical production processes that are part of our daily life.

This work was developed in the context of the research project “QUARREE 100—Resilient, integrated and system-friendly energy supply systems in existing urban districts considering the complete integration of renewable energies” (Grant No. 03SBE113B). The authors thank the project management organization Jülich (PtJ), the Federal Ministry for Economic Affairs and Energy (BMWi), and the Federal Ministry of Education and Research (BMBF). C.S., J.T., and J.I. also want to thank the German Research Foundation (DFG) for funding through the priority program SPP 2080 (Katalysatoren und Reaktoren unter dynamischen Betriebsbedingungen für die Energiespeicherung und -wandlung) under Grant Nos. TH 893/23-1 and BA 1710/31-1. The authors would also like to thank Christoph Hoffmeister (FIBER, Bremen, Germany) and Holger Fricke (Adhesive Bonding Technology department at Fraunhofer IFAM, Bremen, Germany) for their help in the development of the GRP/alumina rods and the workshop of the university for the fabrication of different flanges. Finally, we want to thank Miriam Schubert at BASF Catalyst Germany GmbH for providing the alumina slurry and the used honeycombs.

The data that support the findings of this study are available in the given repository.²⁶ 

Each part of the reactor is linked to its corresponding CAD file²⁶ in Table II alongside the supplier for the part or the used raw material.

The investigated system consisted of two monolithic catalytic honeycombs (cordierite, 600cpsi, BASF Catalyst Germany GmbH, Nienburg, Germany; length ∼ 50 mm and diameter ∼ 20 mm) and non-catalytic front and back heat shields (also honeycombs: length ∼ 25 mm and diameter ∼ 20 mm). Both honeycombs were coated with a 12 wt. % Ru/γ – Al₂ O₃ layer following a coating procedure applied in previous works.¹⁹ To this end, 7.5 g of ruthenium(III) chloride hydrate (41.3 wt. % Ru, ChemPur GmbH, Karlsruhe, Germany) was first dissolved in deionized water and subsequently mixed with 52.5 g of alumina slurry (BASF Catalyst Germany GmbH, Nienburg, Germany). After stirring for ∼5 min to ensure homogeneity of the dispersion, the honeycombs were dipped in the suspension and then cleaned of excess slurry by carefully blowing out the structure using pressurized air. The coated monoliths were then dried at 120 °C for 1 h (heating up 0.33 h) and subsequently calcined at 400 °C for 2 h (heating up 4.4 h). Weighing the monoliths before and after coating and calcination indicates a catalyst loading of 0.33 g and 0.32 g for the first and second honeycombs, respectively.

Catalyst activation was carried out ex-situ using a 3 L_N min⁻¹ (H₂:Ar, 1:4) gas flow at ∼450 °C for 12 h. After cooling down to room temperature (20 °C) in an unchanged gas atmosphere (Ar and H₂), the monoliths were quickly transferred to the NMR reactor, and the reactor was immediately purged with Ar to make sure that no oxidizing species are within the system.

The connection of the peripheral parts can be seen in Fig. 9. For gas flow regulation, mass flow controllers (Bronkhorst Deutschland Nord GmbH, Kamen, Germany) are used with a maximum volume flow of 6 L_N min⁻¹ (FIRC 101-103). The operating pressure is checked through a pressure sensor (PI 201) and a manually operated backflow pressure regulator (Hy-Lok D Vertriebs GmbH, Oyten, Germany). A heating jacket is used to preheat the reactants before entering the reactor. In order to remove water from the product stream, a cold trap (∼−18 °C) was placed at the outlet of the reactor. It was subsequently analyzed using an online process gas analyzer equipped with a non-dispersive infrared detector (GMS810 with a MULTOR module, Sick AG, Germany) for the quantification of CO₂, CO, and CH₄.

A 7 T preclinical NMR imaging system (Biospec 70/20, Bruker Biospin GmbH, Ettlingen, Germany) equipped with a gradient system BGA12S2 (441 mT m⁻¹ maximum gradient strength in each direction and 130 µs rise time) was used for all NMR measurements. A circularly polarized volume RF coil (inner diameter of 72 mm) was used for RF excitation and signal detection. The NMR pulse sequences were implemented using the software platform Paravision 5.1.

The 3D-MRSI sequence was used similar to that described by Ulpts et al.¹⁹ The optimized MRSI pulse sequence used a flip angle of 15°. Immediately after the RF pulse (10 µs rectangular pulse and echo time TE = 0.28 ms), short triangular-shaped phase encoding gradients were applied with circularly reduced k-space sampling. The field of view was (61.5 × 61.5 × 122.5) mm³ in 41 × 41 × 49 voxels (64 × 64 × 49 voxels after zero filling and Fourier transformation). Data acquisition was performed using 512 complex data points and a spectral width of 50 kHz. Using a repetition time (TR) of 12.5 ms, the total time per SI measurement was calculated to be about 8 min. The spatial B₀ homogeneity was optimized by using the standard automatic shimming procedure adjusting only the linear order shims (x, y, z). Shimming was performed prior to measurements of gas reactions using a 40-mm sphere filled with water. For future measurements, we are considering to improve B0 shimming by determining optimized first and second shim currents from MRSI data measured on the gas phase in the reactor. To fit the methane peaks from the measurement results, the matrix pencil method (MPM)³³ was implemented into the post-processing procedure.