A high precision gas flow cell for performing in situ neutron studies of local atomic structure in catalytic materials Free

Catalysts contribute over one trillion dollars to the national gross domestic product in the United States,¹ and yet specifics of chemically reactive atomic environments remain almost entirely out of reach by traditional structural probes due to their aperiodic and transient nature. Current studies require either specialized conditions (e.g., high vacuum²) or sample preparation (e.g., 4-D electron microscopy³). Time-resolved X-ray pair distribution function (PDF) studies have been shown to elucidate changes in the local bonding environment of solids during chemical reactions (without a need for crystalline periodicity);^4–8 yet such X-ray PDF methods offer limited specificity to the chemical constituents or the interface. The emergence of high flux neutron PDF studies provides opportunities to probe oxide, hydride, and other light atom bearing surface species in functional materials,^9–11 with isotopic contrast allowing further precision in chemical specificity through differential PDF analysis.^12–15 At the same time, a powerful technique for revealing kinetic—but not structural—information has evolved with steady-state isotope transient kinetic analysis (SSITKA),¹⁶ involving isotope-exchange experiments. The primary outcome from traditional SSITKA experiments following gas adsorption is the total number of adsorbed molecules in the steady-state, which can be used to derive heats of adsorption and understand the role of chemical composition in reactions and time-dependent gas-specific residencies. More complex kinetic relationships can be elucidated from these methods with the aid of atomistic modeling.

In this contribution, we present a new sample environment developed at the Nanoscale Ordered Materials Diffractometer (NOMAD) beamline¹⁷ at the Spallation Neutron Source (SNS) which allows for precise in situ gas flow reaction studies during neutron total scattering measurements. This gas flow cell allows for studies on small volumes of sample (roughly 1 cm³ of material) which can have a highly controlled quantity of gas flowed through them at specified rates and partial pressures. A high-speed switching valve in the system allows for fast transitions between treatment gases. A residual gas analyzer (RGA) continuously monitors exhaust gases downstream from the sample, and direct encoding of all sample conditions in the data streams allows for time-dependent studies of reaction kinetics such as in situ SSITKA. We present the design elements and an example study on $N2$ adsorption herein, along with a list of anticipated applications.

The gas-flow cell is designed to deliver a specified pressure and flow rate through small samples, with the ability to very quickly transition between two gas species which differ in either chemical composition or neutron scattering power (i.e., different isotopes). The change between these two gases must occur quickly, yet minimize the effect on the physical flow rate and pressure at the sample position. Also critical to the system design is the integration of an RGA posterior to the sample, such that adsorption or reaction processes can be directly measured contemporaneously with diffraction data. The data from the RGA must be encoded in parallel with the diffraction data, such that novel reaction-dependent reduction procedures may be utilized ex post facto.

A design schematic for the gas-flow cell is shown in Figure 1, with associated images of several components shown in Figure 2. NOMAD is a top-loading instrument, and as such, the sample environments are typically designed so that they can be lowered into the standard sample well at the neutron beam position. The high-speed 4-way switching valve (VICI, model ED44UWE) is controlled via a 2-position microelectric valve actuator (VICI, model E2CA), which can be activated manually or automatically with a desired switching frequency (e.g., switching dosing gases every 6 min). In this way, one treatment gas is always flowing towards the sample, while the other is being vented to the system exhaust maintaining a steady-state condition. The flow of these two treatment gases is controlled upstream of the switch by a set of low-pressure drop mass flow controllers (Alicat Scientific, MCW-10SCCM-D). Directly downstream of the switch is a carrier gas port, such that an additional (typically inert) gas may flow continuously with the selected treatment gas. The carrier gas flow is controlled via a separate upstream mass flow controller (Alicat Scientific, MCW-50SCCM-D). Downstream of the sample, a digital back pressure regulator (Alicat Scientific, PC-15PSIG-D) controls the net pressure seen at the sample, independent of the specified flow rates. Note that an identical digital back pressure regulator is installed on the exhaust-side of the high-speed switch, so that any fluctuations in pressure due to the action of physically switching between dosing gases is minimized. Downstream of both the sample and back pressure regulator, but prior to the exhaust vent, an RGA (Stanford Research Systems, UGA200) measures the mass profile of the exhaust gas as it comes off the sample. An optional bubbler system can be installed before exhausting to the atmosphere or the neutron activated gas exhaust system at NOMAD. The entire system is plumbed predominately with 1/16 in. OD, 0.030 in. ID stainless steel tubing, the lengths of which have been minimized to facilitate fast transitions between dosing gas species with regard to the NOMAD geometry. The net length of tubing from the controlling rig to the sample position in the beam and back to the RGA is approximately 6 m. Swagelok fittings were used to connect the majority of the components.

The sample cell itself is a custom made quartz “U-Tube” design utilizing either Ultra-Torr connections (Swagelok, model SS-1-UT-1-2) or stainless steel-to-quartz adapters (Larson Electronic Glass, model SQ-025-T) directly attached to Swagelok valves, such that a sample may be isolated from atmosphere during the mounting and handling procedures. The NOMAD beam is typically collimated to an area of 1.5 cm × 1.5 cm at the sample position. The quartz tube in the sample cell narrows directly prior to the beamspot to the dimensions of OD 5.0 mm and ID 4.62 mm (Wilmad 507-PP-7QTZ), minimizing the background contributions to the diffraction data.

An alternative static gas pressure dosing sample cell is also used, which utilizes the high-precision gas flow instrumentation to expose a sample to a controlled partial pressure of a gas, but does not actively flow the gas through a sample during the diffraction measurement. The sample cell for such a configuration is a sealed quartz tube (Larson Electronic Glass, SQ-018-T) which is connected via a T-section to the flowing gas line between the supplying mass flow controllers and the back pressure regulators. Once a desired gas environment is established, a valve on the sample cell is closed to isolate the sample, thus establishing a static environment.

Under normal operation, the system has a maximum operating pressure at the sample of 15 psig. The two treatment gases can flow up to 10 ml/min, while the carrier gas flows up to 50 ml/min, in both cases with an accuracy of $±0.4%$ of reading +0.2% of full scale. The RGA regularly tracks 4 distinct gas species with a time resolution of 1 s, although more species can be tracked with an associated proportional increase in scan-time. The switching time of the high-speed valve is 425 ms. The sample temperature is controlled via an argon cryostream placed directly below the sample position (Oxford Cryosystems, Cobra Plus) which is capable of running at temperatures of 90 K to 500 K.

Samples to be measured must allow for the free flow of gases at operating pressures, and as such are often pressed and sieved prior to loading to select for the optimal particle size. Due to the relatively small overall volumes associated with the plumbing used, we have found that with a total flow rate of 20 ml/min with 1 psig back-pressure, a switch between treatment gases will be detectable on the RGA within seconds. When a powdered sample is loaded, this response time will increase due to the delay introduced by the gases physically transporting through the sample. This mechanical flow delay is caused by the macroscopic characteristics of the powder composition (sample consistency, sample volume, flow rate, pressure); however, this is separate from and in addition to any chemical interactions or adsorption that may occur between the sample and the flowing gas. In order to decouple any chemical reactivity in the time-dependent response from the mechanical flow delay, an inert gas may be mixed with one of the two dosing gases such that a proper hold-up time can be established during data reduction. For a 3 cm column of powdered sample, which had been pressed into pellets at 4 metric tons for 5 min, broken up and sieved to particle sizes between 100 and 500 microns, the additional delay time at the above conditions was found to be approximately 50 s.

Commissioning experiments with the flow cell were completed on nitrogen adsorption by calcium exchanged zeolite-X. Zeolitic materials are commonly used for air separation by pressure-swing adsorption (PSA)¹⁸ as well as for heterogeneous catalysis such as catalytic cracking¹⁹ and in $NH3$ synthesis as the catalyst²⁰ or as the support material for the catalyst.^21–23 By simultaneously performing in situ neutron total scattering and SSITKA, the experimentally determined structure is linked to reactivity during $N2$ adsorption on the Faujasite cage structure of the zeolite-X.

The Na_78−2xCa_xAl₇₈Si₁₄₄O₃₈₄ samples were produced through ion exchange with sodium zeolite-X (NaX) powder (Sigma-Aldrich molecular sieve 13X) following a conventional ion exchange technique.^24,25 1 g of ground NaX was dispersed in 100 ml of water, into which 1.5 g of calcium chloride dehydrate (EMD Chemicals, Inc.) was added. This mixture was then heated over a boiling water bath for 24 h followed by a filtration step through a Büchner funnel and a washing step with 100 ml of water. The material was dried on the filter at 90 °C overnight. This cation exchange procedure was repeated twice in order to obtain the highest degree of exchange. The sample compositions were characterized using energy dispersive X-ray spectroscopy (EDS) and optical emission spectroscopy with inductively coupled plasma (ICP-OES), both of which found the degree of cation exchange to be nearly complete for Na_78−2xCa_xAl₇₈Si₁₄₄O₃₈₄,x ≈ 38 (CaX), at 98% in the EDS and 97% in the ICP-OES. Laboratory powder X-ray diffraction revealed the samples to be highly crystalline. The morphology of the exchanged sample examined by scanning electron microscopy (SEM) remained unchanged by comparison with those observed for the parent NaX sample. To make the samples more suitable to gas flow measurements, the powders were pressed at 4 metric tons for 5 min, and then broken up and sieved to form a powder with sizes between 100 $μm$ and 500 $μm$⁠. Prior to the neutron total scattering measurements, the samples were charged by drying them under vacuum while heating to 400 °C at a rate of 0.2 °C/min. Samples were stored under vacuum prior to use.

To identify the relevant neutron scattering signatures, static gas loading neutron total scattering measurements were performed at room temperature on NOMAD. Pristine samples under vacuum were measured in the static gas loading configuration for 2 h. The samples were then pressure dosed with $N215$ flowing at 5 ml/min. The back-pressure at the sample was cycled between 1 and 5 psig to ensure complete exposure of the sample and dilution of any present contaminant gas (e.g., $N214$⁠). The samples were sealed at 1 psig and measured for 2 h at 300 K. The measured and fit diffraction patterns from these measurements are shown in Figure 3.

Rietveld analysis of the data was performed using TOPAS refinement software,²⁶ in which the Ca exchanged zeolite-X structure model was fit to both the pristine and nitrogen pressure dosed sample diffraction data. For the pristine data, the composition of the material was used to set occupancy values, with the exception of calcium. The calcium sites are considered not fully occupied, and additionally the remaining unexchanged sodium would affect the refined calcium occupancy due to the difference in scattering power. Calcium sites (1) and (2) are inversely related in occupancy, as these two sites cannot be occupied simultaneously. In fitting the nitrogen dosed data, the occupancy of all atoms except nitrogen was fixed to values found in the pristine case. We found that the occupancy of the nitrogen is strongly correlated to the atomic displacement parameter (B_iso), and therefore we fixed the occupancy of nitrogen to the value found by the in situ SSITKA (see Sec. III C). The refined parameters from these fits are summarized in Table I. Higher quality refinements were found possible in the pressure dosed sample case through the inclusion of nitrogen atoms near the Ca(3) positions, which were not required in fitting the pristine data. The presented fits in Figure 3 include data from the NOMAD detector bank centered at 154° $2θ$ angles.

The volume of the zeolite-X unit cell clearly contracts upon exposure to the nitrogen by approximately 2.3%, as would be expected during adsorption.²⁷ Furthermore, the fit occupancy of nitrogen in the pressure dosed case correlates to approximately 8 N-atoms per unit cell, well above the value expected if only free flowing gas was present (⁠$≈0.4$ N-atomsper unit cell based on ideal gas law calculations). The results of this refinement demonstrate that neutron total scattering techniques are sensitive to the presence of nitrogen adsorption at pressures as low as 1 psig and room temperature, well below the pressures (⁠$CD4$ (9 bars) and $C2H6$ (5 bars)²⁸) and above the temperatures (Xe (210 K),²⁹ $CD4$ (77 K),³⁰ $CHCl3$ (20 K),³¹ or $CO2$ (4 K)³²) used in previous gas loading measurements.

The total scattering and resultant PDF datasets from the static loading measurements of the pristine and pressure dosed samples on NOMAD are shown in Figure 4. Clear differences can be observed after the introduction of nitrogen to the system. In particular, the first peak in the difference curve of the PDF (Figure 4(b), red) at $≈1.1Å$ suggests sensitivity to the $N2$ pairwise molecular bond. The observed differences require further modeling to interpret.

The flow cell design allows for time-dependent kinetic studies of samples under steady-state conditions as well as pump-probe gas flow. To demonstrate this capability, we performed in situ SSITKA measurements with simultaneous neutron total scattering. Here, the two treatment gases were $N214$ and $N215$⁠, where the $N214$ was mixed with 3% Ar to act as an inert tracer gas in the SSITKA measurement to determine the gas-phase hold-up time. Helium was used as an inert carrier gas. Approximately 1 cm³ of the fully charged material, prepared as described above, was loaded into the gas flow cell. Steady-state flow and pulse-probed SSITKA were performed, in each case with 1 ml/min treatment gas and 20 ml/min carrier gas with the pressure at the sample held at 1 psig. During the SSITKA measurements, the switching time between treatment gases was set to 6 min and the complete cycle between the two isotopes was performed 12 times for a total measurement time of $≈150$ min; testing indicated 6 min allowed for a complete changeover from one isotope of nitrogen to the other. The averaged results of the SSITKA measurements are shown in Figure 5(a).

The two most informative parameters determined in a SSITKA experiment are the surface concentration of adsorbed nitrogen atoms, N, and the average surface residence time, τ. The latter is calculated by integrating the area between the normalized transient response of the $N214$ isotope and the normalized transient response of the inert tracer argon.³³ The surface concentration of the nitrogen is determined by multiplying the average surface residence time by the total flow rate in the system.³³ The number of nitrogen molecules per unit cell was determined based on the surface concentration. The results of the SSITKA were found to be comparable to the offline commercial SSITKA measurements (Altamira Instruments AMI-300TKA) performed on the samples as shown in Figure 5(b) and Table II. Note that although the measured nitrogen adsorption per unit cell was very similar between the offline and in situ studies, the observed differences can be attributed to subtle variations in sample preparation and activation procedures, strongly justifying the use of simultaneous in situ SSITKA with diffraction.

As this information is recorded directly into the metadata of the NeXus files³⁴ used for powder diffraction data reduction, we utilize the Mantid data analysis framework³⁵ to reduce and analyze the scattering data via advanced event-based methodologies.³⁶ Such a comprehensive integration of software and hardware allows for novel uses of the gas flow sample environment, such as pulse probe studies, looping, and SSITKA, even after the experiment is completed.

The stroboscopic nature of the gas isotope measurement during the in situ SSITKA experiments provides neutron total scattering data for specific gas isotope conditions as a function of time. The results of the neutron total scattering measurements were found to be consistent with static gas loading results, but inadequate for quantitatively differentiating between the isotopic species of nitrogen gas. A full utilization of the data may be possible with longer measurement times and advanced stroboscopic data reduction methods employing in-line RGA data.

A high-precision gas flow cell sample environment has been developed for neutron total scattering measurements at the SNS. Combined with the intense neutron flux available at the NOMAD instrument, novel opportunities to directly address the interactions between gas-solid interfaces are now possible. A commissioning experiment on N₂ adsorption in zeolite-X combined in situ neutron diffraction and SSITKA measurements. Of important significance, the effect on the overall lattice structure and the adsorption location of N₂ were identified at 300 K and 1 atm, in line with operating conditions of the material, and in striking contrast to the conditions of previous studies. The gas flow cell with in-line RGA, neutron diffraction, and neutron PDF capabilities is a versatile sample environment that will aid in identifying structural responses, characterizing nanoscale interfaces, and following structure transformations in chemical, biological, and geological gas-solid processes alike. Steady-state and time-dependent phenomena can be explored as a function of temperature, pressure, or flowing gas species. A wide range of fundamental and applied science uses is anticipated, including the identification and evolution of chemical compositions and unique atomic structure environments involved in heterogeneous catalysis, gas storage, oxidation/reduction processes, solid-oxide fuel cell cycles, sensor technology, capture of atmospheric pollutants, ion exchange processes, and more.

In the future, the flow cell’s capability to deliver time-dependent isotope contrast to reacting/growing material interfaces in the neutron beam can be developed, together with stroboscopic data binning, to probe the kinetics of detailed surface structure species in steady-state processes (for example, utilizing the contrast of ¹⁴N versus ¹⁵N for improved sensitivity to N₂ adsorption/fixation, NH₃ synthesis, or NO reduction processes, or utilizing the contrast of ¹²C versus ¹³C for improved sensitivity to CH₄ oxidation, CO oxidation, propene epoxidation, or CO₂ reduction/fixation). Extensions to liquid flow in this setup can enable studies on liquid-solid material processes, including studies of environmental effects on medical implants and cycling behavior in electrochemical insertion electrodes and batteries. Overall, this new sample environment will provide new opportunities to validate theory and simulation, demonstrate synthesis and operation, and elucidate structural underpinnings of behavior in a host of gas-solid material phenomena.

This work was primarily supported under the Department of Energy’s Office of Basic Energy Sciences and the Laboratory Directed Research and Development (LDRD) Program at Oak Ridge National Laboratory (LDRD Seed No. 7735). The presented analysis of neutron powder diffraction data was funded by the BES Early Career Award: Exploiting Small Signatures: Quantifying Nanoscale Structure and Behavior KC04062, under Contract No. DE-AC05-00OR22725. The data presented were measured on the Nanoscale-Ordered Materials Diffractometer (NOMAD) instrument at the Spallation Neutron Source at ORNL. Some samples were prepared and additional characterization was performed at the Center for Nanophase Materials Sciences, which is a DOE Office of Science User Facility. The authors acknowledge and thank Genevieve Martin for contributing the photography used in Figure 2.