Reactivity of Bioinspired Magnesium-Organic Networks under CO2 and O2 Exposure.

Photosynthesis is the model system for energy conversion. It uses CO2 as a starting reactant to convert solar energy into chemical energy, i.e., organic molecules or biomass. The first and rate-determining step of this cycle is the immobilization and activation of CO2, catalyzed by RuBisCO enzyme, the most abundant protein on earth. Here, we propose a strategy to develop novel biomimetic two-dimensional (2D) nanostructures for CO2 adsorption at room temperature by reductionist mimicking of the Mg-carboxylate RuBisCO active site. We present a method to synthesize a 2D surface-supported system based on Mg2+ centers stabilized by a carboxylate environment and track their structural dynamics and reactivity under either CO2 or O2 exposure at room temperature. The CO2 molecules adsorb temporarily on the Mg2+ centers, producing a charge imbalance that catalyzes a phase transition into a different configuration, whereas O2 adsorbs on the Mg2+ center, giving rise to a distortion in the metal-organic bonds that eventually leads to the collapse of the structure. The combination of bioinspired synthesis and surface reactivity studies demonstrated here for Mg-based 2D ionic networks holds promise for the development of new catalysts that can work at room temperature.

Introduction

For almost 3 billion years, nature has developed fantastic methods to convert solar energy to fuels in a way that humans are trying to replicate. This process is called photosynthesis,[1] which occurs in two steps: the light reactions (which require light) and carbon fixation (also known as the dark reaction or Calvin cycle). Up to now, there exists only a single conversion technology with prospects for long-term large-scale use: photovoltaics.[2,3] In a photovoltaic cell, photons are absorbed and converted to electrical energy (light reaction). This leads to electric currents that have to be used immediately. Pursuing the long-term goal of direct fuel production from solar energy requires mimicking the other part of photosynthesis, conversion of CO2 (dark reaction).

However, activating CO2 and, more importantly, converting it to fuels is notoriously difficult due to the high stability of the CO2 molecule, which imposes significant energy demands and requires high-temperature/pressure conditions and/or active reductants, such as hydrogen.[4−6] Alternatively, solid-state catalysts are needed. The vast majority of the autotrophs in nature fixate the inorganic carbon atom from CO2 to produce sugars, during the dark reaction of photosynthesis, with an enzyme as catalyst: the Ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO, Figure a). This macroprotein is the most abundant protein in the biosphere. The active center of the RuBisCO is a Mg2+ ion surrounded by amino acid residues whose functions are to anchor the ribulose and provide the CO2 as a cofactor bound to a lysil residue in the surroundings of the Mg2+ (Figure a).[7] At lower levels of CO2, O2 molecules compete with CO2. This competition consumes extra energy and, eventually, liberates CO2. Thus, the enzyme is considered inefficient and motivates scientists to investigate ways to improve the enzymatic function of RuBisCO.

Figure 1

(a) Model of the enzyme RuBisCO and its active center.[40] (b) Scanning tunneling microscopy (STM) image of a typical terephthalic acid (TPA) network on Cu(100) at room temperature (RT). The inset shows the structural model of the fully deprotonated molecules arranged in a typical 3 × 3 configuration (Vbias = 1.122 V; i = 70 pA). (c) An STM image showing the structure formed by the co-deposition of TPA molecules and Mg atoms (Vbias = 0.821 V; i = 50 pA). The TPA molecules are characterized by the circular motifs whereas Mg, represented in the model by green circles, appears as circular protrusions between three TPA molecules. NB: the apparent height (brightness) of the Mg atoms depends on the tip conditions (see Figure S2).

Metal–organic networks (MONs) are structures that resemble well the basic configuration of metalloenzymes.[8] They consist of metal nodes bound to organic bridging ligands that self-assemble on a metal substrate. These nanoarchitectures hold great potential in heterocatalysis due to the tunability of their metal centers,[9−11] organic linker, pore size,[12,13] and physical/chemical properties.[14−18] Even though many MONs were reported, the study of their catalytic properties with respect to gas adsorption is still in its infancy. Nevertheless, we consider that there is a critical mass of knowledge to attempt the inversion of the structure–function relationship. Our approach is then to apply the nanoarchitectonic principles learnt in the past two decades to start building structures that could resemble the architecture of the active site of known enzymes.[8,19] This approach promises to provide a playground where an atomistic insight into the relevant processes can be achieved, and, ultimately, it will allow us to create new bioinspired catalysts that could work at room temperature (RT). Moreover, the high level of control of the coordination environment allows us to evaluate separately the contribution of each functional group to the reactivity of the center.

Molecules with carboxylic functional groups (R-COOH) can work as excellent building blocks in the design of molecular nanostructures.[20] One characteristic is the deprotonation of COOH into a carboxylate group (COO–) on solid surfaces at RT, which confers them the ability to form robust metal–organic networks with co-deposited transition[17,18,21−24] as well as alkali metal atoms.[25−28] Terephthalic acid (1,4-benzenedicarboxylic acid, TPA) is a model molecule that easily deprotonates on Cu surfaces at RT and eventually forms metal–organic networks with co-deposited adatoms.[17] A second characteristic is in regard to the oxygen atoms from its COO– anions acting in the networks as electron acceptors,[29] which allow the control of the redox potential in the metal centers.[17,30] This particular feature encourages the use of metal–organic networks for O2 dissociation,[8,30,31] aimed to resemble the catalytic function of enzymes specialized in O2 reduction.[32,33]

The adsorption of CO2 was recently explored on 2D metal–organic networks. One of the few examples involves the use of metal–organic chains of 1,4-phenylene diisocyanide molecules bound to Au adatoms on Au(111), to promote CO2 capture at 77 K.[34,35] Moreover, networks formed by molecules with carboxylate groups bound to Fe centers have shown adsorption of CO2 at 100 K.[36] Nevertheless, so far, no CO2-driven phase transition has been reported at RT for any 2D metal–organic array.

Herein, inspired by the structure of the active site in the RuBisCO enzyme, we present a method to replicate the carboxylate environment around a Mg2+ ion. For this purpose, we designed an ionic network with Mg2+ as active centers by the self-assembly of TPA molecules with Mg atoms deposited on Cu(100) at RT. Finally, in the attempt to test the catalytic function of these bioinspired complexes, high partial pressures of CO2 and O2 are exposed on the system at RT. The stability and chemical changes in the network related to gas adsorption are studied by complementary techniques such as scanning tunneling microscopy (STM), high-resolution X-ray photoelectron spectroscopy (HR-XPS), and first principles density functional theory (DFT) calculations.

Results and Discussion

Formation of Mg-TPA Networks on Cu(100)

The main goal of this work is to mimic the catalytic function of the active site of the RuBisCO enzyme through the fabrication of 2D ionic networks with Mg2+ centers. Surface science studies have demonstrated that the catalytic activity of metal particles is boosted when they are reduced to single atoms,[37] as the coordination number decreases. Furthermore, organic molecules can coordinate adatoms from a metal surface, controlling the chemical environment around single atoms and their distribution, leaving apical orbitals ready to react with new species. This characteristic makes them good candidates to study their potential in heterogeneous catalysis. The first step in the strategy toward the formation of ionic networks with Mg adatoms consists in the formation of a self-assembled layer of the organic linkers. TPA molecules (coverage = 0.4 ML) were deposited on an atomically clean Cu(100) surface at RT. Earlier studies have reported that TPA molecules deprotonate and self-assemble on a Cu surface at RT.[38,39] Indeed, the typical STM image of this system (Figure b) shows a homomolecular patch of TPA forming a 3 × 3 phase structure (with respect to the underlying Cu surface) related to the self-assembly of fully deprotonated molecules.[38,39]

The second step consists in the co-deposition of sub-ML coverages of Mg atoms and TPA molecules on an atomically clean Cu(100) surface at RT (Figure c). The deposition process was finalized with postannealing at 353 K for 5 min to favor the diffusion of the species on the surface. STM images (Figures c and S1) reveal the formation of a new structure incorporating single Mg atoms coordinated to three TPA molecules. Larger area images also show the presence of larger Mg clusters (bright circular protrusions) surrounded by TPA molecules (Figure S1-a). As we shall see, the presence of Mg clusters also agrees with XPS analysis in the Mg 2p region. Figure c shows that the molecules in the ordered islands get grouped in consecutive trigonal-planar arrangements, suggesting an electrostatic coordination to Mg2+. The model superposed (obtained by DFT calculations) presents a structure based on a Mg2+ ion, represented as a green solid circle, coordinated by three TPA molecules through their carboxylate moieties. Two types of Mg–carboxylate coordination are present: one TPA with a bidentate carboxylate coupling to Mg, and two TPA carboxylates interacting via a monodentate coordination with Mg. From now on, this structure will be referred as Mg-3TPA. To the best of our knowledge, this is the first 2D metal–organic architecture that incorporates an alkaline earth metal atom (group 2) as a metal center.

A more precise configuration of the Mg-3TPA structure is described by DFT calculations. The calculations were performed using the Vienna ab initio simulation package (VASP),[41,42] including van der Waals (vdW) interaction within the vdW-dF scheme. Due to the large unit cell and the consequential computational overcharge, one monolayer of the Cu(100) surface with Cu atoms in fixed positions was included in the calculations, whereas the rest of the atoms forming the organic overlayer were allowed to relax until forces were smaller than 0.05 eV/Å. Tests including two substrate layers indicate a negligible (no more than a 2%) change in the vertical distances as compared with the calculations with a monolayer of the Cu substrate (see Figure S3 and Table S1). As a result, a relaxed Mg-3TPA ionic network was obtained on top of a Cu(100) layer. A Bader charge analysis[43,44] revealed the ionic character of Mg and O atoms due to the rearrangement of charge in the system. As seen in Figure a, for modeling the surface, a (10 × 5) unit cell was considered and the overlayer consists of two Mg atoms that lay at almost bridge positions on top of the surface, each of them coordinated with four O atoms from three different TPA molecules, i.e., six TPA molecules in total. The vertical distances from the atoms of the overlayer to the Cu surface vary from 1.96 Å the closest to 3.40 Å the largest. The two TPA units that are aligned with the [100] crystallographic direction get tilted around 7° and are anchored to the surface by the non-coordinated COO– group, whereas the other two diagonal TPA units adopt an arched conformation. In agreement with the crystallographic database for the average bond distance in Mg–carboxylate networks,[45] the O–Mg distances for the monodentate and bidentate coordination are 2.03 and 2.20 Å, respectively. In addition, each TPA molecule forms H bonds (O···H–C) between two of their uncoordinated carboxylate moieties and the aromatic rings from adjacent molecules. To visualize the flux of charge due to the formation of the metal–organic compound, the electronic charge density difference is depicted in Figure b. The remarkable charge depletion (red) on the Mg positions and the charge accumulation (blue) in the O atoms that interact with the Mg adatom observed in Figure b represent a fingerprint of the ionic character of the complex formed. To confirm the reliability of the model, the Tersoff–Hamann approach[46,47] was used to simulate STM images of a typical Mg-3TPA network (Figure c). The simulation was based on the integration of spatially resolved density of states in energy, from a certain applied bias potential (−2.0 eV) to the Fermi level. This simulation reproduced the zigzagging inversely ordered trigonal-planar arrangements seen in the experimental STM images of the Mg-3TPA network (see the combined image at the center of Figure c).

Figure 2

(a) Top and side views of the relaxed configuration of the ionic network on top of a Cu(100) layer calculated by means of DFT. The preferential geometry is based on Mg2+ ion coordinating three TPA molecules. The unit cell consists of two Mg2+ ions; each cation coordinates four carboxylate oxygens from three different TPA molecules. (b) Electronic density difference of the optimized structure showing the accumulation (blue) and depletion (red) of charge. Contour: 0.005 e–. (c) Combined simulated STM image (top and bottom) of the system using the Tersoff–Hamann Approach with a bias voltage of −2 V and an experimental STM image (center).

The chemical identity of TPA (Figure a) and Mg-3TPA (Figure b) networks on Cu(100) was confirmed by HR-XPS spectra. The XPS spectrum in the C 1s region of the reference system shows the typical profile of a homomolecular network containing fully deprotonated TPA molecules: an intense peak at 284.6 eV is attributed to the carbon atoms from the phenyl ring (6C), and a small peak at 287.6 eV, characteristic of carboxylate groups (−COO–).[38] The dotted peak at 285.1 eV could be related to the different chemical environment around the aromatic carbons.[38] The 3 × 3 structure is also confirmed by the analysis of the XPS spectrum in the O 1s region, which contains a single peak at 531.3 eV attributed to carboxylate oxygens (−COO–).

Figure 3

(a) HR-XPS spectra in the C 1s and O 1s binding energy regions of the homomolecular 3 × 3 phase of fully deprotonated TPA species on Cu(100). (b) HR-XPS spectra for a typical ionic network prepared by the co-deposition of TPA molecules and Mg atoms on Cu(100). The core level shifts in the C 1s and O 1s regions evidence the electrostatic interaction between Mg2+ and TPA species.

The shifts of the peaks in every spectrum in Figure b suggest an interaction between the co-deposited TPA molecules and Mg adatoms. On the one hand, the main peaks of the C 1s and O 1s binding energy regions in the network formed by the co-deposition of TPA molecules and Mg adatoms on Cu(100) (Figure b) are shifted toward higher binding energies, compared to those obtained for the TPA homomolecular network. Similar core level shifts were already reported as related to TPA molecules acquiring a protonated character on Cu(100).[38] However, TPA molecules are fully deprotonated under the present deposition conditions. Therefore, the protonated character of the negatively charged COO– groups would be endowed by the Mg2+ cations, thus evidencing the formation of ionic networks. The detailed analysis in the C 1s region reveals a small peak at higher binding energies (289.4 eV), attributed to the electrostatic interaction between the COO– group and the Mg2+ cations and an intense single peak at 285.2 eV, attributed to the 6C ring of the fully deprotonated TPA species interacting with Mg2+. The remarkable core level shift to higher binding energies is associated to the charge migration from the C atoms toward the oxygen atoms in COO–.[25] In addition, the O 1s line shape shows a dominant peak at 532.1 eV, related to a relative less negatively charged COO– group due to the low electronic density around the core electrons during the electrostatic COO––Mg2+ interaction.[25] On the other hand, at lower energies, the Mg 2p binding energy region shows a broad peak that can be split in two components: one feature at 50.8 eV that reveals the presence of small Mg clusters on the surface and a second peak at 51.4 eV in the region corresponding to the ionic form of Mg (Mg2+).[48−51] These two peaks suggest that TPA molecules sequestrate Mg atoms from clusters to eventually form ionic networks with Mg2+ cations. This mechanism may occur in a similar way as seen in ionic networks formed with alkali compounds.[52]

So far, we have described the formation of a 2D structure (Mg-3TPA) after the deposition of TPA molecules and Mg adatoms on Cu(100). This structure is stabilized by the ionic interactions between the carboxylate groups from TPA and the Mg2+ centers and resembles the carboxylate environment of the Mg2+ ion present in the active site of the RuBisCO enzyme. Next, we explore the catalytic properties of these ionic networks for gas adsorption.

Reactivity of 3 × 3 TPA and Mg-3TPA Ionic Networks with CO2 and O2

To study the reactivity of these assemblies, the 3 × 3 TPA and the Mg-3TPA adsorbed on Cu(100) structures were exposed to 470 L of CO2 or 450 L of O2. In the case of the homomolecular TPA network, STM images did not reveal any structural change neither when exposed to CO2 nor O2, preserving the 3 × 3 structure intact after gas exposure (Figure S4). Thus, we conclude that the pure TPA network does not interact either with CO2 or with O2. In fact, the molecular network passivates and protects the Cu surface for oxidation, as oxygen rows are only observed to grow in the bare Cu regions. To further verify the lack of chemical changes that could be undetectable by STM, HR-XPS experiments were also done. As shown in Figure S5, CO2 is not absorbed in bare Cu regions at RT, as it cannot be detected in the C 1s and O 1s regions. Moreover, the 6C and COO– signals in the C 1s region did not show any variation compared with the sample before gas exposure (Figure S5), confirming that CO2 is not adsorbed in the molecular phase. Conversely, the exposure to O2 molecules generates a peak at 529.7 eV characteristic of O–2 species chemisorbed on bare Cu surfaces,[53] whereas the peak at 531.3 eV remains at the same energy, confirming that the COO– groups do not react with O2. These experiments confirm that the TPA molecular phase is unreactive under CO2 or O2 exposure and that O2 rows grow in the bare Cu areas, as seen by STM (S4).

This control experiment is important as proof of the formation of a reactive metal–organic network for two main reasons. First, the use of the same conditions to deposit TPA on Cu(100) at RT guarantees the full deprotonation of the molecules because carboxylate oxygens bind and control the redox potential of the metallic centers, and second, the unreactive character of COO– groups against CO2 and O2 is crucial to discard the possible role of these groups as active centers when they self-assemble with Mg adatoms.

A typical ionic network with a Mg-3TPA configuration was prepared using the standard preparation process stated above. Figure a shows a series of consecutive STM images taken on a specific area of the surface over different periods of time. As seen in the first three images, the pristine Mg-3TPA structure was tracked during a period of 67 min and no structural changes were observed. Thus, the Mg-3TPA network was found suitable to test its reactivity under these conditions. Afterward, the interaction with CO2 was tracked during almost 2 h. The first image (marked with a black border, snapshot with t = 0 min) represents the moment when CO2 is released on the system. After only exposing 22 L of CO2 (10 min), the surface area of the network decreases, providing the first evidence of reactivity. Then, after 99 L of CO2 exposure (22 min), noticeable structural changes appear, related to a significant decrease of the Mg-3TPA network domain at the expense of a new phase forming at the bottom right side of the snapshot. The new phase presents a different configuration and it turns predominant after 256 L of CO2 (57 min). This square-shaped structure organizes the molecules in a new motif: one Mg adatom surrounded by four TPA molecules leading to the formation of pores. From here on, this structure will be referred as Mg-4TPA. This new square-symmetrical arrangement suggests one type of Mg–carboxylate coordination: monodentate couplings (O–Mg) of the Mg center with each TPA molecule.[17] In addition, each TPA molecule forms H bonds (O···H–C) between two of their uncoordinated carboxylate moieties and the aromatic rings from adjacent molecules (Figure b). The pores in the network are the result of the convergence of four uncoordinated carboxylate oxygens from four TPA molecules. After 473 L of CO2 (105 min), the coverage of the predominant Mg-4TPA phase increases whereas the former Mg-3TPA structure vanishes completely. Finally, the CO2 injection was stopped and the images were tracked during the next 17 min to observe whether or not the structure could reverse from a Mg-4TPA to Mg-3TPA configuration. The last snapshot at 122 min shows the stability of the Mg-4TPA structure with no signs of structural reversibility. Therefore, this experiment demonstrated that CO2 catalyzed a phase transition: from a Mg-3TPA to Mg-4TPA configuration. This phase was also observed at high TPA/Mg ratios; in fact, Figure S6 shows pure TPA domains coexisting with the Mg-4TPA phase.

Figure 4

(a) Snapshots taken from a time-lapsed sequence of STM images for a Mg-3TPA ionic network on Cu(100) before, during (images with a black border), and after being exposed to CO2. An irreversible structural change is observed: the Mg-3TPA network evolves to a Mg-4TPA structure (Vbias = −1.057 V; i = 73 pA). (b) Sketch of the CO2–mediated transition from Mg-3TPA to Mg-4TPA. The middle panel shows a top and a side view of the Mg-3TPA model used to evaluate the local CO2 adsorption. DFT calculations show that the intermediate between one structure and the other is based on the adsorption of a linear CO2 molecule on the reactive Mg2+ center.

The structural changes observed by STM images encouraged the use of DFT calculations to study the local charge distribution in detail when CO2 interacts with the ionic networks. Figure b shows that the presence of CO2 distorts the local bonding in a Mg-3TPA structure. CO2 adsorbs on the Mg2+ center in a linear configuration related to a weak chemisorption, with charge transfer weaker than that in a bent sp2 configuration that is normally related to CO2 chemisorption,[36] which requires the formation of the anion. In our DFT calculations, the adsorption energy of CO2 on the Mg-3TPA network is −0.69 eV. This does not correspond to a full chemisorption, but this interaction seems to be strong enough to induce weakening of the bonds between Mg2+ and the carboxylate oxygens from TPA, disrupting the structure and inducing the binding of a fourth TPA to Mg2+ to form the Mg-4TPA structure (as seen after 22 min in Figure a). Many thorough studies have been related to the CO2 adsorption on low-index Cu surfaces. On Cu(100), Cu(110), and Cu(111), CO2 is physisorbed in a linear configuration at rather low temperature (and desorbed at 70–85 K) with adsorption energies in the ranges of 0.22–0.31 eV.[54,55] In a former paper, Campbell et. al.[56] indicate that the chemisorption of CO2 on a clean Mg(0001) single crystal is dissociative, resulting in both surface oxide and carbonate species. Nevertheless, as we shall see from our XPS spectra, no signs of carbonates and magnesium oxides were detected during our experiments. Therefore, we conclude that the calculated adsorption energy of CO2 on the Mg-3TPA network (more than twice the adsorption energy of CO2 on bare Cu surfaces) is enough to produce an electronic rearrangement on the network, inducing a reconfiguration that triggers the formation of a new structure. However, no evidence of CO2 dissociative adsorption is detected.

A second experiment was done to study the interaction of the ionic networks also with O2 molecules. Again, consecutive STM images were taken on a defined area on the surface over different periods of time (Figure a). The first snapshot revealed the presence of the Mg-3TPA network and this ionic network did not show any structural changes during an interval of 30 min. The snapshots highlighted using a black border correspond to consecutive STM images taken during the O2 exposition for a period of up to 2 h. The first structural change is observed after the exposure of 27 L of O2 (6 min). The ionic network begins to disassemble from its center and this tendency continues while more O2 is released in the system. Hence, after 171 L of O2 (38 min) is exposed, most of the Mg-3TPA structure is disassembled. The disassembly is more evident after exposure of 342 L of O2 (76 min), whereas the borders of the terraces also present strong signals of oxidation. Finally, a fully disordered structure is observed after exposure of 567 L of O2 (126 min). Contrary to the case of CO2 exposure, the structure never shows reorganization into an ordered configuration. These observations probe the direct interaction of O2 with the system, which induces complete and irreversible disruption of the Mg-3TPA phase. Due to the disruption of the Mg-3TPA phase, an incipient formation of the 3 × 3 TPA phase is also observed together with the disordered phase (S7-a).

Figure 5

(a) Snapshots from a time-lapsed sequence of STM images for a TPA-Mg ionic network on Cu(100) taken before and during (images with a black border) O2 exposure, where the Mg-3TPA evolves to a disordered molecular phase (Vbias = 1.189 V; i = 74 pA). (b) The interaction of O2 induces the transition from Mg-3TPA to a disordered phase. The middle panel shows the intermediate structure of the local adsorption of O2 obtained by DFT calculations.

As described previously, CO2 catalyzes phase transition from Mg-3TPA to Mg-4TPA. In contrast, the phase transition of Mg-3TPA to a disordered structure is rather specific to O2. To study the local structure when O2 interacts with the Mg-3TPA network, we performed DFT calculations showing the interactions of O2 molecules on a Mg-3TPA network. The results showed that O2 also binds nondissociatively through one of its oxygen atoms to the Mg2+ center of the ionic network (Figure b). The absorption energy of O2 on top of a Mg2+ ion in the Mg-3TPA network was calculated to be −0.79 eV.

The interaction of a metal in an O2-rich environment leads to the formation of a metal oxide that implies the dissociation of O2. The Cu(001) surface is the most reactive among the low-index Cu surfaces to the adsorption and direct dissociation of O2, as evidenced at various temperatures (even at 5 K).[57] On the other hand, the Mg(0001) surface also undergoes oxidation at RT, leading to the formation of MgO thin films under exposure of O2 doses.[51,58] Nevertheless, it was also demonstrated by first principles calculations that O2 molecules can be incorporated into Mg adatoms by nondissociative adsorption to produce MgO2 admolecules on a MgO(001) surface with point defects.[59] Thus, it is possible that after O2 exposure, the uncovered Cu surface will undergo oxidation (by dissociation of O2 molecules) and different oxidized Mg species will be formed in the network by dissociative and/or nondissociative interactions. Figure S7-b shows how O2 molecules react with the Cu substrate (only when the surface is not fully covered by the Mg-3TPA network) and form the well-known oxygen rows on Cu(100).[60] Despite the fact that the DFT calculations are not conclusive (because the absorption energy difference between the physisorption of CO2 and O2 in the network is lower than expected), the calculated adsorption energy at the first step of the O2 interaction is in good agreement with the trend observed experimentally. Therefore, the slightly higher adsorption energy obtained for O2 seems to be enough to generate a reconfiguration of the charge around Mg2+ ions and to destabilize the COO––Mg2+ coordination leading to the dissociation of the metal–organic network.

The chemistry of CO2 and O2 interaction with the ionic networks was again contrasted using HR-XPS (Figure ). As a reference, we included the Mg-3TPA ionic network spectra before gas exposure (Figure a). After 945 L of CO2 exposure, XPS spectra do not show any significant change in the Mg 2p, C 1s, and O 1s binding energy regions (Figure b) compared to those in the Mg-3TPA network before exposure. As discussed above, CO2 physisorbs on the Mg2+ center and it catalyzes a phase transition from a Mg-3TPA into Mg-4TPA structure. Therefore, the slight difference in the COO––Mg2+ coordination between both structures may produce changes in the chemical environment around Mg and O atoms. The Mg 2p region, shows an increment in the relative peak area at 51.4 eV (COO––Mg2+) at the expense of the peak at 50.8 eV (Mg cluster with few atoms). This fact probes the action of CO2 to enhance the segregation of the remaining small Mg clusters into Mg adatoms, which eventually coordinate with deprotonated TPA molecules to form larger patches of ionic network. Indeed, the segregation of the Mg clusters can be related to the joint action of CO2 and deprotonated TPA molecules. It was shown above that fully deprotonated TPA species can segregate the clusters into single atoms of Mg, whose distribution on the surface is controlled by the COO––Mg2+ coordination. Thus, in contrast with ref (51), Mg atoms in the ionic network promote only a weak physisorption of CO2, preventing the formation of MgCO3.

Figure 6

HR-XPS spectra showing the C 1s, O 1s, and Mg 2p binding energy regions for a typical Mg-3TPA network before (a) and after being exposed to 945 L of CO2 (b) and 500 L of O2 (c), respectively.

Concerning the peaks related to C atoms from carboxylate (COO–) and phenyl (6C) groups from the TPA molecules that form COO––Mg2+ coordination, they remain at 289.4 and 285.2 eV, respectively. Hence, the presence of physisorbed (if any) species of CO2 is imperceptible. The presence of absorbed CO2 is also elusive in the O 1s region (Figure b). The difficulty to identify the peaks related to CO2 leads considering also the possibility that CO2 may have left the system. Nevertheless, the consequences of the interaction of CO2 on the system could be detected in the O 1s region where the peak at 532.2 eV (Mg2+–COO–) is slightly shifted by about 0.1 eV to higher binding energies. This shift could be related to the different coordination environment of the COO– oxygens present in the recently formed Mg-4TPA configuration.

Summing up, even though CO2 species cannot be observed within the time resolution of our XPS measurements, STM images and DFT calculations showed that the interaction of CO2 is strong enough to reorganize the charge around Mg2+, hence triggering a transition from Mg-3TPA to Mg-4TPA coordination. Thus, the ensemble of the results obtained by STM, DFT, and XPS suggests that CO2 interacts with the system through temporary physisorption on Mg2+.

The chemistry of O2 adsorption in the ionic networks was also studied by HR-XPS. Figure c shows the Mg 2p, C 1s, and O 1s binding energy region of a Mg-3TPA network after 500 L of O2 exposure. At first sight, it can be observed that the main peaks are shifted to lower binding energies compared with the Mg-3TPA network before O2 exposure. The O 1s binding energy region shows an intense peak at lower binding energy (529.7 eV) related to the presence of oxidized Cu, as a consequence of the reactivity of O2 with the bare Cu(100) areas, as confirmed also by the oxygen rows present in the STM image in Figure S7-b. However, the reactivity of O2 with the ionic network is evidenced by the broader peak arising at a higher binding energy, whose deconvolution reveals two peaks at 531.1 and 531.6 eV. As studied by STM and DFT above, O2 quenches the charge in the Mg2+ center and consequently disrupts the COO––Mg2+ interactions. In fact, the peak at 531.1 eV is characteristic of the COO– group of deprotonated TPA molecules (see Figures SI and S2)[38] and confirms the disassembling of the TPA molecules from the ionic network. Accordingly, the peak at higher binding energy resembles the peak characteristic of the oxidized Mg species.[58]

The disassembling of the TPA molecules from the network is also confirmed in the C 1s region. The prominent peak at lower binding energies splits into two peaks: 284.6 and 285.2 eV, respectively. The peak at 284.6 eV is attributed to C6-ring carbons in fully deprotonated TPA molecules.[38] This peak presents a higher relative area compared with the smaller peak at 285.2 eV characteristic of the ionic networks, whose presence evidences the remaining small areas with the COO––Mg2+ coordination. Hence, the broad peak at higher binding energies at 288.4 eV is attributed to the wide range of contributions between the COO– groups from the free molecules surrounded in a disordered fashion. Finally, the Mg 2p region, shows a broad peak at 50.6 eV attributed to Mg-oxidized species. Its shift to a lower binding energy would result from the effect of CuO on the substrate,[61] in a similar way as seen with larger Mg clusters exposed to O2. These results demonstrated that the O2 interaction induces the quenching of the charge without a complete breakdown of the COO––Mg2+ coordination. This implies that TPA molecules remain in a disordered configuration but they do not completely segregate to form a homomolecular network (in agreement with STM images, Figure a).

Conclusions

Inspired by the structure of the active site of RuBisCO, we have presented a method to synthesize a system based on a Mg2+ center stabilized by a carboxylate environment. Similar to the RuBisCO enzyme, the catalytic activity of the system was tested under high relative pressures of CO2 and O2 at RT. The Mg2+ centers were provided by the evaporation of Mg atoms on the surface. The full deprotonation of TPA molecules on Cu(100) at RT has been found suitable to provide the carboxylate environment around the center of Mg2+. The co-deposition of TPA molecules with Mg atoms on Cu(100) at RT resulted in the formation of a 2D ionic network. The structure was stabilized by the electrostatic interaction between Mg2+ and TPA species. The unit cell is formed by a single Mg2+ cation surrounded by three deprotonated TPA molecules (Mg-3TPA). The ionic networks have shown structural evidences of gas adsorption upon exposure to relative high-pressure doses of CO2 or O2. Both CO2 and O2 were found to interact with the Mg+2 center. The CO2 molecules adsorb temporarily on Mg2+, producing a charge imbalance that induced a phase transition into a different configuration (Mg-4TPA), whereas O2 adsorbs on the Mg2+ center, giving rise to a distortion in the metal–organic bonds that eventually leads to the collapse of the structure. In both cases, the observed phase transitions showed that the ionic networks favor gas adsorption at RT. The 2D ionic network presented herein shows the potential of metal–organic coordination on surfaces to mimic enzymes active sites and track structural changes in situ under gas exposure by scanning probe microscopy, providing insight on the surface reactivity of coordinated metal centers and holding promise for the design of heterogeneous catalysts tailored down to the atomic level.

Experimental Section

STM experiments were performed in a homemade scanning tunneling microscope operated under ultra-high vacuum (UHV) conditions and repeated in the Omicron STM available at the synchrotron. This is an important experimental check to make sure that the networks we measured with XPS are the same measured by STM. The XPS data were acquired in the X-Ray beamline PEARL synchrotron facilities at the Paul Scherrer Institute (PSI), Switzerland. The data for the Mg 2p, C 1s, and O 1s were acquired with a beam energy of 160, 370, and 650 eV, respectively. The pass energy of the analyzer was set to 20 eV. The spectra were referenced using the Au 4f7/2 line at 84.1 eV. All experiments were carried out under UHV (∼ 3 × 10–10 mbar) at RT.

Substrate Preparation

A Cu(001) single crystal (Mateck) is used as a substrate. The cleaning process was performed in the UHV chamber with a base pressure of 5.0 × 10–10 mbar. The surface was cleaned by consecutive cycles of Ar+ sputtering (0.5 keV) and annealing at temperatures up to 810 K, followed by a slow and gradual cooling down of the crystal to 313 K before starting with the next cleaning cycle.

Sample Preparation

1,4-Benzenedicarboxylic acid (TPA) molecules (Fluka) were deposited by thermal evaporation from a quartz crucible by heating it up to 448 K. During molecular deposition, the pressure in the chamber was kept below 2.0 × 10–9 mbar while the temperature of the substrates was kept at RT. Before deposition, the molecules were fully degassed.

Computational Details

DFT calculations were carried out using the 5.4.4 version of VASP.[42,62,63] For the description of electron–ion interactions the projector-augmented wave method was employed, whereas the Perdew, Burke, and Ernzerhof functional was used to describe exchange and correlation within the generalized gradient approximation[64] and optB88-vdW functional to describe the nonlocal van der Waals interaction.[65]

The periodic slab models include a Cu layer, the two nonequivalent Mg-3(4)TPA units, and 15 Å of vacuum. A 2 × 2 × 1 Monkhorst-Pack mesh is used to describe the unit cell, and the plane-wave energy cutoff was 400 eV in all cases. To perform the relaxation of the Mg–organic network on top of the Cu (001) surface, all atoms of the network were relaxed until the forces were less than 0.05 eV/Å and a convergence criterion of 10–4 eV was used for energy. However, when the small reactants were included, only the latter and the corresponding Mg center and the surrounding carboxylate groups were allowed to relax.

VESTA software[66] was used to represent the geometrical and electronically induced density models with output data of the previous calculations. The STM simulation of the adsorbed Mg-TPA on Cu(100) substrate in the Tersoff–Hamann approach was done with WSxM software.[67]