Multiscale investigation of adsorption properties of novel 3D printed UTSA-16 structures

facilitated by the development of a new non-aqueous ink formulation, employing hydroxypropyl cellulose and boehmite to adjust the rheology of the ink. What makes this formulation and printing process di ﬀ erent from the printed adsorbents and catalysts published previously, is that the resulting structures in this work were not sintered. The presence of the binder matrix not only produced the physical properties for printability but also ensured a homogeneous dispersion of UTSA-16 in the structures, as well as gas adsorption characteristics. The monoliths were tested for the adsorption of di ﬀ erent gases (N 2 , CH 4 , CO 2 and H 2 O) in order to apply them into separation processes that contribute to defossilizing energy and fuels production. Water is strongly adsorbed in this material (~14 mol/kg at 293 K) and is competing with CO 2 for adsorption sites. Breakthrough curves showed that the retention time of CO 2 decreases signi ﬁ cantly when the feed stream is saturated with water. In this study, synchrotron XRD-CT data were collected in situ, in a non-destructive way, and phase distribution maps were reconstructed to, for the ﬁ rst time, gain insight into the spatial and temporal evolution of the UTSA-16 containing phases in the operating 3D printed monolith during the exposure to CO 2 .

• XRD-CT operando insight on chemical and structural variation within 3D printed MOF.

Keywords:
Carbon dioxide Water adsorption Direct write 3D printing Shaping In situ operando XRD-CT A B S T R A C T Structuring MOF materials is a fundamental step towards their commercialization. Herein we report intensive characterization of 3D-printed UTSA-16 monoliths, facilitated by the development of a new non-aqueous ink formulation, employing hydroxypropyl cellulose and boehmite to adjust the rheology of the ink. What makes this formulation and printing process different from the printed adsorbents and catalysts published previously, is that the resulting structures in this work were not sintered. The presence of the binder matrix not only produced the physical properties for printability but also ensured a homogeneous dispersion of UTSA-16 in the structures, as well as gas adsorption characteristics. The monoliths were tested for the adsorption of different gases (N 2 , CH 4 , CO 2 and H 2 O) in order to apply them into separation processes that contribute to defossilizing energy and fuels production. Water is strongly adsorbed in this material (~14 mol/kg at 293 K) and is competing with CO 2 for adsorption sites. Breakthrough curves showed that the retention time of CO 2 decreases significantly when the feed stream is saturated with water. In this study, synchrotron XRD-CT data were collected in situ, in a nondestructive way, and phase distribution maps were reconstructed to, for the first time, gain insight into the spatial and temporal evolution of the UTSA-16 containing phases in the operating 3D printed monolith during the exposure to CO 2 .

Introduction
Metal Organic Frameworks (MOFs) are considered ideal candidate materials for innovative gas separation and adsorption applications involving small gas molecules such as CO 2 and CH 4 [1][2][3][4][5]. Their large internal surface area and the high porosity built up from organic molecules (ligands) and metal nodes or clusters contribute to very large adsorption capacities [6,7].
Adsorption processes are already a mature commercial technology for hydrogen purification and for biogas upgrading. If these technologies are integrated in a system implementing selective CO 2 capture, they can contribute to defossilizing energy and fuel production respectively, mitigating the effect of CO 2 in climate change. An example of successful integration of the existing H 2 -PSA has been achieved in Port Arthur refinery. MOF materials applied to these separations can have a significant impact in footprint reduction resulting in improved process economics.
Metal-organic framework UTSA-16, K(H 2 O) 2 Co 3 (cit)(Hcit) (Hcit = citric acid), has been shown to possess one of the highest loadings of CO 2 per volume of adsorbent [30][31][32]. There are two main reasons for this: -The favourable interactions between CO 2 and disordered water sites occupying large pores in the material (each K + ion coordinates two structural water molecule sites at ambient pressure) [31,33,34]. -The density of UTSA-16 crystals is 1.659 g/cm 3 and the particle density is close to zeolites, which means that the weight of UTSA-16 packed in a fixed column volume is higher than other MOFs [11].
To produce shaped structures of UTSA-16 it is essential to ensure that the printing suspensions contain at least 50% of alcohol (ethanol, propanol, etc) in all steps of the production [11]. The difficulty in the shaping is that such alcohols have a very different evaporation rate than water. Most recently, stable performance and relatively good adsorption kinetics of 3D-printed UTSA-16 monoliths with 85 wt% MOF content was reported [26]; a CO 2 adsorption capacity of 1.31 mmol/g was obtained, which is equivalent to 87% of the powder capacity under the same conditions (5000 ppm of CO 2 at 25°C).
In the current work, UTSA-16 monoliths produced by 3D printing were fully investigated for their use within adsorption processes. The material was characterized by X-Ray diffraction, thermo-gravimetric analysis and scanning electron microscopy. More insight into the crystal structure and spatial and temporal distribution across the adsorbing 3D printed UTSA-16 monolith was gained by performing in situ high-energy X-ray Diffraction Computed Tomography (XRD-CT) studies. The present study is the first of its kind to explore adsorption in an operating 3D printed MOF monolith combining the use of in situ/operando XRD-CT under a CO 2 flow. In a previous synchrotron operando study by the authors, XRD-CT was employed to provide information on the spatial distribution of the active phase in a working 3D printed catalyst monolith for CO 2 methanation [35]. The 5D XRD-CT imaging technique, combined with new fast tools for simultaneous data collection and analysis allow for the chemistry inside working reactors to be imaged, assessing the fidelity of (3D printed) material design and determine the chemical and physical form of active components [36]. In terms of applications prospects, we measured the adsorption equilibrium of CO 2 , CH 4 , N 2 and H 2 O at three different temperatures (293, 313 and 343 K). We also measured breakthrough curves of CO 2 -N 2 and CO 2 -N 2 -H 2 O mixtures.

Experimental
UTSA-16 was synthesized following a previously reported recipe [30,31]. The hydrothermal ynthesis was done using the following compounds: Co(OAc) 2 , 4H 2 O,C 6 H 8 O 7 , H 2 O, KOH, H 2 O and C 2 H 5 OH, which were put in a Teflon-lined 5L Parr autoclave reactor in a molar ratio of 1:1:3:139:43. The reactor was heated to 393 K and kept at that temperature for 2 days before letting it cool down to ambient temperature. Over 100 g of UTSA-16 powder was produced in each batch using these molar ratios.
The formulation of the ink in this work was based on non-aqueous solvent to avoid possible interaction of water molecules with the UTSA-16′s citrate ligands for prolonged periods of time. The preparation of the printing ink of the UTSA-16 polymer composites was carried out using the tuneable rheology of an adequate selection of precursors: 45 wt% of a 6% hydroxypropyl cellulose (HPC) solution with isopropyl alcohol (IPA) and 18 wt% nanosized Boehmite AlO(OH) based dispersant.
The printing formulation has been developed by taking into account the effect of blending parameters (such as the stepwise addition of solid precursors, rotational speed in a planetary mixer and constant ambient temperature) to ensure sufficient dispersion of the active UTSA-16 material throughout the polymer micro-lattices.
The direct-write 3D printing of the multi-material ink (consisting of active UTSA-16, with HPC/IPA solution and AlO(OH) as binders) involved its micro extrusion and fibre deposition through a nozzle by print head movement in xyz. The UTSA-16 containing ink was printed into 3D periodic ('log pile') cylindrical structures with a diameter of 28 mm using a high-end nScrypt 3D-450HP printer (see Fig. 1). The printed cylinders were subsequently slowly dried in a controlled  humidity environment and at ambient temperature. The final monoliths all have the same final composition (Table S1 in ESI summarises the dry solids content of the monoliths). The current adsorbent monolith preparation method differs from the ones that were previously described [37,38] in that the UTSA-16-loaded polymer matrix was not removedno thermal treatment was applied allowing the entire composite cylinder to densify into a solid bulk to achieve the final shape. The thermal stability of the UTSA-16 monoliths was analysed by using a TG-DSC-MS instrument (Netzsch STA 449 F1 instrument equipped with a QMS 403C MS analyser). Experiments were carried out using approximately 30 mg pieces of the monolith, 2 K/min heating rate and a nitrogen flow of 50 ml/min. The samples were heated to 1073 K. Mass changes (TG), heat transfer within the sample (DSC) as well as analyses of the kind of volatile components leaving the sample during the process (by mass spectrometry -MS) were recorded.
Subsequently, XRD-CT measurements were performed on a 3D printed UTSA-16 monolith after exposure to pure CO 2 at a flow rate of 100 ml/min. The XRD-CT measurements were carried out at the ID15A beamline at ESRF using a monochromatic beam with an energy of 90 keV focused to a spot size of 40 μm × 20 μm (Horizontal × Vertical). XRD-CT scans were collected using a zigzag approach over a total angular range of 180°in steps of 0.721°(i.e. 250 angular steps) combined with translational steps of 40 μm (with a total number of 320 steps across 12.8 mm) using a Pilatus3 X CdTe 2 M detector. Each complete XRD-CT scan comprised 80,000 diffraction patterns. The total acquisition time per point was 20 msec and each XRD-CT scan lasted for a total of ca. 20 min. The detector calibration was performed using a CeO 2 NIST standard. Every 2D diffraction image was converted to a 1D powder diffraction pattern after applying a 10% trimmed mean filter to remove outliers using the nDTomo and pyFAI software packages [39][40][41][42]. The data integration was performed with a fast GPU processor. The final XRD-CT images (i.e. reconstructed data volume) were obtained using the filtered back projection algorithm; see the resulting tomographic reconstructions in Figure X corresponding to peak intensity, position and FWHM constructed after peak fitting for the first three diffraction reflections (1 0 1), (1 1 2) and (2 0 0). To obtain a clearer view of the changes in the three peak parameters the ratio of these maps was calculated, and histograms plotted for each individual peak.
Conventional lab X-ray diffraction (XRD) measurements were performed using a Philips/Panalytical X'Pert Pro diffractometer set up in Bragg-Brentano geometry with Cu Kα source with λ = 1.5418 Ǻ at 40 kV and 40 mA at room temperature. Powder X-Ray diffraction patterns from 2ϴ = 5 to 50 of the MOF adsorbent were measured using a PANalytical EMPYREAN diffractometer equipped with a Cu-source and Pixel 3D detector.
To determine microstructure and composition, Scanning Electron Microscopy (SEM) and Energy Dispersive X-ray (EDX) spectroscopy images were obtained using a FEI Nova NanoSEM 450 field emission gun (FEG), operated at an accelerating voltage of 25 keV, and fitted with a Bruker QUANTAX 200 EDX system and XFlash 5030 SDD detector.
Adsorption isotherms of pure gases were measured in a Belsorp Max manometric unit. Nitrogen isotherms at 77 K were used for the estimation of the BET surface area. Isotherms for CO 2 , CH 4 and N 2 were measured at 293, 313 and 343 K up to 101 kPa. For water, measurements were carried out until its saturation partial pressure or to a maximum of 4 kPa. The adsorbent was degassed overnight at 393 K under vacuum before the measurements. Adsorption equilibrium data was fitted with the Virial isotherm. A theoretical description of the isotherm and parameters used for the fitting are given in the Supplementary information.
Breakthrough curves of the binary (CO 2 -N 2 ) and ternary (H 2 O-CO 2 and N 2 ) mixtures were measured in a standard unit for measuring dynamic experiments. A more detailed description is given in the Supplementary Information. For that purpose, two 3D-printed structures (each with 19 mm diameter and 9 mm length and weight of 3.7103 g) were introduced into a one-inch stainless steel tube. Prior to the experiments, the material was degassed at 393 K under a constant flow of helium that was also passed through the adsorbent before starting the breakthrough curves. The experiments were made with a total flowrate of 100 ml/min composed by 20% CO 2 and 80% N 2 . For the experiment with water, a saturator was used, and the flowrate and gas composition was maintained (same molar fraction in dry basis). The real molar fractions are affected by the 2.2% of water introduced in the experiments.

Adsorbent characterization
What makes this formulation and printing process different from those published previously is that the resulting structures in this work were not sintered, and the sacrificial material was not removed. Typically, 'green' printed structures are subjected to debinding and sintering processes to completely combust the organic binders and free up sorption sites, ensuring the process results in the least possible effect on the final structure integrity. The current work avoided any additional processing steps preserving the inter-particle cohesion and mechanical robustness while keeping a reduction in the surface area to a minimum. While retaining the functionality, the present binder matrix not only facilitated physical properties for printability but also enabled gas diffusion and adsorption characteristics.
3D printed monoliths maintained a porous structure after solidification and a considerably higher BET surface area of 540 m 2 /g. Table  S2 in ESI provides a summary of the Nitrogen adsorption measurements at 77 K and obtained pore size profiles. The measured isotherm for a 3D printed monolith sample can be found in Fig. S1. The nitrogen adsorption measurements on the 3D printed monoliths produced Type IIisotherm curves during the initial stage, which was subsequently, however, similar to the Type IV isotherm containing a hysteresis loop of Type H4 [43]. The hysteresis loop relates to a marked increase in N 2 adsorption over a relative pressure range of 0.4-0.8. While Type II is the most common isotherm for nonporous and macroporous adsorbents, Type IV is indicative of the presence of mesoporous structures that greatly vary in shape and are associated with subtle changes in adsorption and desorption behaviour. In addition, the H4 hysteresis corresponds to capillary condensation in the mesopores that are slit-shaped and, in terms of size, closer to the micropore region.
SEM images in Fig. 2 show the macro-and micro-structure of 3D printed UTSA-16 containing monoliths prior to the adsorption measurements. HPC polymer binder and Boehmite-based additives were employed as support materials in order to tailor the plasticity and mechanical properties of the resulting 3D printed composite monoliths as well as their physico-chemical functionality (diffusion pathways and surface properties). Elemental EDX point analysis and EDX maps reveal UTSA-16 constituents (Co and K species) uniformly dispersed within the binder matrix and discrete regions of Boehmite (AlO(OH) features and particles of up to 20 µm in size) across the 3D printed fibres. For elemental composition analysis, see the individual EDX spectra from across the representative regions of the 3D printed fibre in Fig. S2 in ESI.
Figs. S3 and S4 in ESI present ex-situ lab XRD data on the original UTSA-16 powder as well as 3D printed UTSA-16 monoliths before and after CO 2 adsorption. The unit cell parameters correspond to those obtained in the literature (13.0691(4) Å b 13.0691(4) Å c 30.1570(10) Å) [34]. The changes of the a, b and c parameters in the orthorhombic phase before and after CO 2 adsorption are shown in Table S3 (ESI).
The in situ XRD-CT measurements confirmed the crystal structure and unit cell parameters' changes in the bulk material upon adsorption, measured by the lab XRD instrument (See Figs. S5 and S6 in ESI for further details). This was followed by a detailed comparison of the spatial distribution of peak intensities, peak position and FWHM, measured across a horizontal cross section in the 3D printed UTSA-16 monolith using high energy X-ray diffraction at the ID15A beamline. On introducing 100 ml/min CO 2 into the tubular reactor, structural changes were observed for all three main reflections (101, 112 and 200). The combined data on the phase distribution maps were constructed to provide more detailed information on physical or chemical spatial variations. The distribution maps in Fig. 3 feature some spatial (local) variations in concentrations of all three phases and changes in peak intensity, peak position and FWHM. All the three sets of in situ XRD-CT data suggested that the overall behaviour is consistent across all three reflections revealing that the 3D printed UTSA-16 adsorbent evolved under CO 2 conditions. There is an apparent increase in peak intensity and peak positions. The shifts in 2θ peak point to changes in lattice parameters following adsorption, availability of adsorption sites and the possible presence of intercalated H 2 O molecules. The trend in these changes are evidenced by the ratios of intensities of the parameters before and after the exposure to CO 2 for the three main reflections (see relative changes in histograms in Fig. S6).

Thermal stability of UTSA-16 monolith
The TG-DSC-MS traces during heating at a heating rate of 2°C/min to 800°C is shown in Fig. 4. Below 100°C, adsorbed water is released, shown by an endothermic DSC signal and an increase in the MS water signal. At around 100°C, the TG trace reaches a flat plateau indicating that most water has been desorbed before the mass trace starts falling between 150 and 200°C. This first reduction in mass is caused by a part of the samples being combusted by the traces of molecular oxygen within the instrument: a small exothermal peak in the DSC curve is accompanied by a sharp reduction in the O 2 level and an increase in the CO 2 level. At around 330°C there is a sharp endothermic peak caused by the total decomposition of the material. This peak is accompanied by sharp releases of CO 2 , H 2 O and also small amounts of O 2 . After the decomposition, the mass trace levels off, flattening out at 53.1% of the initial mass. Assuming that UTSA-16 decomposes to the metal oxides (CoO and K 2 O), the amount of Al-oxide dispersant can be estimated at around 18 wt%.

Adsorption equilibrium
The adsorption equilibrium of N 2 , CH 4 , CO 2 and H 2 O measured at Fig. 3. Structural evolution and active sites distribution across the polymer matrix of 3D printed monolith with peak intensity, peak position and FWHM spatial distribution across a representative cross section for three main UTSA-16 reflections (200, 101, 112) before and after CO 2 adsorption has taken place.   293, 313 and 343 K is shown in Figs. 5 and 6. Measurements were carried out on two samples (denoted as S1 and S2) coming from different monoliths produced with the same batch. The results indicate that the difference of amount adsorbed of carbon dioxide in the two samples is small and less pronounced in the Henry zone of the isotherms. The amount of nitrogen adsorbed is very small and is linear within the whole region. Methane is adsorbed stronger than nitrogen and the isotherms are only slightly non-linear at pressures closer to atmospheric. Carbon dioxide is more selectively adsorbed reaching a loading of 2.5 mol/kg at 298 K and atmospheric pressure. Water is the most strongly adsorbed gas reaching a capacity of 14 mol/kg at 293 K at almost 100% relative humidity (corresponding to 2.1 kPa). Moreover, the water isotherms are not Type I according to IUPAC classification. Since they present multiple inflection points in different locations depending on the temperature used for the measurements, it is necessary to use many parameters to describe them. The solid lines in the figures are the fitting of the Virial model. It can be seen that the level of complexity of the model depends on the shape of the isotherms; there are such a high number of parameters in the description of the water isotherms, that it is probably not correct to extrapolate the data. The capacity of CO 2 and water is around 60% compared to the reported data on the UTSA-16 powder [32]. Reduction of capacity and surface area was also reported after shaping [44,45]. Comparing UTSA-16 with other commercial materials, the CO 2 capacity is intermediate, ranging between that of zeolite 13X and activated carbon [46][47][48]. The main advantage of UTSA-16 is that the isotherms of CO 2 are not as steep as those of zeolite 13X which can result in less power consumption for regeneration.

Breakthrough curves
To show how the samples will perform with real mixtures, we have measured binary and ternary breakthrough curves. In Fig. 7, the results of a CO 2 -N 2 breakthrough curve is displayed. It is seen that the sample is able to selectively retain CO 2 over nitrogen. The shape of the curve indicates a rather steep breakthrough of CO 2 which shows that dispersion and diffusional effects are not large. Moreover, there is a longer tail in the curve that is frequently caused by thermal effects; the adsorption of CO 2 is exothermic which results in a net increase of the temperature of the adsorbent reducing its CO 2 capacity. While the adsorbent is saturated at the local (slightly higher) temperature, the gas starts to cool down increasing its capacity and adsorbing more CO 2 until the equilibration at the right temperature is achieved. The shape of the breakthrough does not show any kind of anomalous shape due to the defects in the 3D printing of the material.
The results of the ternary breakthrough water of H 2 O-CO 2 -N 2 are presented in Fig. 8. When water is present in the feed stream, the breakthrough time of CO 2 is comparable to the one in the dry experiments once that the water concentration front moves much slower than the CO 2 front. However, the CO 2 plateau moves back to the feed concentration once that water gets adsorbed indicating that CO 2 is displaced by water. The experiments were performed twice and no difference in the results were observed. It has been hypothesised by Masala et al [31], that CO 2 adsorption can be attributed to K + species, forming K + (CO 2 ) and K + (CO 2 ) 2 adductsrather than that being a water-directed interaction. There were initial concerns that working with the feed stream containing close to 100% relative humidity could cause the structure of the material to be damaged, but according to the measurements, the experiments were reproducible, and no material degradation or deactivation was observed.

Conclusions
We report the development of a new non-aqueous cellulose-based ink formulation, and successful 3D-printing of UTSA-16 composite monoliths. The rheology of the ink used for printing was adjusted by adding an appropriate quantity of hydroxypropyl cellulose and boehmite. The monoliths contained 18% of boehmite. The surface area of the 3D-printed sample is~540 m 2 /g indicating a reduction of 30% when compared to the powder used. Gas diffusion pathways through   the microstructure of the monoliths were not limited by the presence of the polymer matrix. Adsorption of N 2 , CH 4 , CO 2 and H 2 O were measured at three different temperatures up to atmospheric pressure (or 4 kPa for water). The isotherms of nitrogen are linear in the 0-100 kPa region, while methane is slightly nonlinear, with a small amount adsorbed (0.3 mol/kg at 293 K and 100 kPa). Carbon dioxide loading is 2.5 mol/kg at 100 kPa and 293 K and 1 mol/kg at 10 kPa indicating that the isotherm is not very steep. This is crucial for reducing the power consumption of regeneration in adsorption processes. The amount of water at 293 K at saturation pressure is 14 mol/kg and the isotherms of water are not Type I IUPAC classification. The breakthrough curves showed that the diffusion of gases is fast and water can easily displace carbon dioxide. This confirms that water competes with CO 2 for adsorption sites. The 3D-printed monolith was used for several breakthrough measurements with water and there was no degradation or reduction of capacity observed during the experimental campaign.
In situ synchrotron XRD-CT was employed to gain complementary information pertinent to the spatial and temporal evolution of the UTSA-16 phases in the 3D printed monolith during adsorption under a flow of CO 2 at ambient pressure. The XRD-CT data were analysed and plotted to reconstruct phase distribution maps of each parameter (diffracted intensity of selected peak area, peak position, and peak width) across a representative slice in the 3D printed monolith. A complete understanding of the structure response and distribution of the active phases under operating conditions is crucial to providing scope for the further optimisation of adsorbent design, properties and performance.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.