Aberration-Corrected STEM to Determine the Surface Coverage and Distribution of Immobilized Molecular Complexes

doi:10.21203/rs.3.rs-3568894/v1

Download PDF

Article

Aberration-Corrected STEM to Determine the Surface Coverage and Distribution of Immobilized Molecular Complexes

https://doi.org/10.21203/rs.3.rs-3568894/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The surface immobilization of molecular catalysts is attractive because it combines the benefits of homogeneous and heterogeneous catalysis. However, determining the surface coverage and the distribution of a molecular catalyst on a solid support is often challenging, inhibiting our ability to control catalytic performance. Here, we demonstrate that scanning transmission electron microscopy can image the location of the metal center in surface-attached transition metal complexes with atomic resolution. Using a machine learning model, we can analyze many images to determine surface coverage and distribution in a non-destructive manner. This allows us to establish how changes to the molecular catalyst affect surface coverage and distribution. Our work describes a new method to characterize surface-attached catalysts, which is likely general to many systems.

Physical sciences/Nanoscience and technology/Techniques and instrumentation/Microscopy/Transmission electron microscopy

Physical sciences/Chemistry/Catalysis/Catalyst synthesis

Physical sciences/Chemistry/Green chemistry/Renewable energy

Physical sciences/Chemistry/Electrochemistry/Electrocatalysis

The immobilization of molecular transition metal catalysts on solid supports has long been attractive because the hybrid material combines many advantages of homogeneous and heterogeneous catalysis^1-7. The molecular nature of the catalyst allows tuning of catalytic activity and selectivity by modifying the metal or ancillary ligands, while the heterogeneous nature has the advantage of facile separation from products. However, the characterization of supported catalysts has been challenging, which limits reproducibility, control of catalytic performance, and the design of new systems⁸. The characterization of immobilized molecular catalysts has typically relied on infrared, photoelectron, and x-ray spectroscopies⁹. These techniques provide essential information about the surface species, but determining the surface coverage of a molecular catalyst is its own challenge. In some cases, electrochemistry can be used to determine surface loading when catalysts are attached to highly conductive supports and give well-behaved cyclic voltammograms. These conditions are not met on semiconductors that could serve as light-absorbers¹⁰. Destructive techniques such as inductively coupled plasma mass spectrometry (ICP-MS) can provide overall surface coverages if all the catalyst is successfully removed from the surface. However, none of these techniques give insight into the location and distribution of the catalysts, features that are often directly related to catalytic performance. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) is a potential method to locate molecular catalysts on the surface via the detection of the transition metal¹¹. This technique has a sub-Ångstrom resolution, high sensitivity to atomic number, and a small depth of field¹². It also does not require a high substrate conductivity, although it does require a very thin sample (~ 30 nm) to provide the required electron transmission. Previously, HAADF-STEM has been used to image single metal atoms or metal oxide catalysts^13-18, but it has not been used to image immobilized molecular catalysts with tunable organic ligands. These types of catalysts bind to surfaces differently than metal oxide or single-atom systems and can facilitate a much wider array of transformations.

Immobilized molecular catalysts on semiconductors are appealing for photo-electrocatalytic applications, such as CO₂reduction^19-23. This is because (i) selectivity, which a molecular system can improve, is essential in converting CO₂ to valuable products, and (ii) rapid transport of electrons from the light-absorbing support to the catalyst can provide high current densities. Silicon (Si) is a highly abundant semiconductor capable of absorbing visible light and transferring photogenerated electrons to a molecular catalyst on the surface^24-29. Here, we use HAADF-STEM to image the metal centers in immobilized molecular catalysts on Si supports used for either CO₂ reduction or the hydrogen evolution reaction (HER). We measure the surface coverage and quantify the catalyst distribution by locating the heavy metal atoms in the molecules. The heavy metal atoms are easily discernible in the images because they scatter considerably more electrons than the supporting materials (Fig. 1b)¹². To obtain values that represent the entire sample, we analyze many images recorded at different places with convolutional neural network (CNN) models^30,31. The clear visibility of the metal atoms and the automated atomic detection from many images allow us to gain insight into the arrangement of the immobilized catalysts and provide a statistical estimation for the entire sample. Further, by varying the immobilization conditions, we study how the reaction conditions used for catalyst immobilization affect catalyst distribution and loadings. Our results show that the combination of HAADF-STEM and deep learning is a valuable and potentially non-destructive method for characterizing immobilized catalytic systems.

A significant challenge for imaging surface-attached molecular catalysts using HAADF-STEM is the need for immobilization on a thin support. Besides its desirable light-absorbing properties, Si is an ideal support due to the availability of thin Si wafers. In this work, we used commercially available microfabricated Si TEM grids (from SiMPore Inc.) that contain nine ‘Si windows’. The Si windows are sections of 100 oriented n-type Si and are approximately 100 µm wide, 100 µm high, and 30 nm thick (Fig. 1a). The Si TEM grids were either etched in buffered HF (or (NH₄)HF₂) to terminate the surface with hydrogen atoms, or coated with a layer of TiO₂ (5 nm) deposited via atomic layer deposition (ALD). The Si grids containing the TiO₂ layer were thinned to 20 nm using a reactive ion etcher before the deposition to maintain electron transparency. We initially immobilized three different third-row, transition metal-based molecular catalysts for CO₂ reduction or HER on Si-based substrates, as shown in Fig. 1c. The complexes Re{4‑(3‑buten‑1‑yl)‑4'‑methyl‑2,2'‑bipyridine}(CO)₃Cl (Re-^Bubpy)³² and Re{5‑(4‑vinylphenyl)‑1,10-phenanthroline}(CO)₃Cl (Re-Phen)³³, which contain a bipyridine or phenanthroline ligand with a pendant alkene, were covalently attached to hydrogen-terminated Si-substrates using a hydrosilylation reaction that forms Si–C bonds³⁴. The hydrosilylation reactions were promoted thermally (160 °C) and performed for 1 hour for Re-^Bubpy and 15 minutes for Re-Phen, respectively under an N₂ atmosphere to prevent oxidation. The Pt porphyrin complex, Pt(4‑(10,15,20-triphenylporphyrin‑5‑yl)benzoate) (Pt-Porph)^35,36, which contains carboxylic acid groups on the porphyrin, was immobilized on a TiO₂-coated Si substrate through a hydrolysis reaction performed for 1 hour (Fig. 1c).

Representative HAADF-STEM images are shown in Fig. 1d-f for Re-^Bubpy, Re-Phen, and Pt-Porph, respectively. Below each image is data from x-ray photoelectron spectroscopy (XPS) of a control Si wafer (100, degenerate n-type, 380 mm thickness) that was functionalized in an analogous fashion with the TEM grids. We propose that the bright spots in the images, which are sufficiently resolved to locate their positions, represent individual metal atoms associated with the molecular complexes. They were observed on both sides of the Si windows for Re-^Bubpy, and Re-Phen, but we mainly analyzed images from the bottom of the thin film because these had less contamination. In contrast, in the case of Pt-Porph, we imaged the top side of the Si because the TiO₂ layer was only deposited there (Fig. 1f and Supplementary Fig. 2a). In the images of Pt-Porph, we also observed some lower contrast spots, presumably representing the Ti atoms. This is consistent with the fact that Ti has an atomic number between Si and Pt, and those lower contrast spots were not observed from the bottom side of the window (Supplementary Fig. 2b) or from the images of the Re complexes (Fig. 1d-e). For the Re complexes, XPS shows most of the Re is in the +1 oxidation state, consistent with the molecular complexes being present and intact on the surface after immobilization. The small amount of Re(0) and an even smaller amount of ReO₃present only for Re-^Bubpy^37-39implies some decomposition and air-induced oxidation, respectively, during or after the immobilization. In the case of Pt-Porph, XPS shows only Pt(II) on the surface. The surface coverage and the distribution of the molecules were evaluated from more than 15 images, each with a field of view (FOV) measuring 16.4 nm × 16.4 nm, for each sample. Utilizing a literature framework (30), two CNN models trained by more than 20 images of Re-^Bubpy or Pt-Porph, each having a field of view (FOV) measuring 8.2 nm × 8.2 nm. These models effectively detected individual atoms in each image (See Supplementary Note 1). The average surface coverage of Re-Phen (3.13 atoms/nm²) was higher than that of Re-^Bubpy (1.61 atoms/nm²), even though the time for the hydrosilylation reaction for Re-Phen was a quarter of Re-^Bubpy (Fig. 1g). The surface coverage of Pt-Porph on the TiO₂ overlayer was 0.21 atoms/nm². The small surface coverage of Pt-Porph suggests that hydrosilylation is a more efficient strategy for achieving high surface coverages than hydrolysis with a carboxylic acid on TiO₂.

A key question relating to the immobilization of catalysts on a surface is whether the molecules are randomly distributed or whether there is a preference for some aggregation on the surface. This type of question is almost impossible to answer using techniques such as cyclic voltammetry or XPS. We analyzed our images using a spatially descriptive statistical method, Ripley’s K-function, to quantify the distribution of the molecules⁴⁰ (See Methods). To simplify the analysis, we calculated K_dev(r) by subtracting the K-function of complete spatial randomness from the K-function of the experimental image. In a completely random distribution, K_dev(r) is expected to be perfectly flat at a value of zero, while positive or negative K_dev(r) values indicate clustering or dispersion at a specific distance, respectively.

We calculated K_dev(r) from more than 15 images for each sample and averaged them (Fig. 1h-j). In the cases of Re-^Bubpy and Re-Phen, there is a negative deviation at small distances (~ 2.0 Å, marked with blue arrows). This is not surprising as the heavy metal atoms in these complexes cannot be this close together. There also is a positive peak at around 4 Å (marked by red arrows). This indicates that the images show several clusters of about 4 Å radius. A likely reason for this is that there was some decomposition of the molecules and formation of Re(0) clusters with Re-Re bonds⁴¹. Consistent with this hypothesis, XPS indicates that the Re(0)/Re(I) ratio is 0.2:1 and 0.13:1 for Re-^Bubpy and Re-Phen, respectively (Table S1 – S2). The K_dev(r) of Pt-Porph showed no peaks until 15 Å, indicating no clustering of the molecules. Presumably, there is no dip at close distances in this case because the surface coverage is low. In all samples, a slight increase in the K-function as the radius increased from 0 Å to 15 Å was observed, implying that there are preferable nano-scale regions for molecules to be bound to the surface. We attribute this observation to surface conditions such as irregular, air-induced oxidation or roughness in deposition.

The surface coverage and distribution of a molecular catalyst on a support are likely key to catalytic performance, but without information about these parameters, it is difficult to control them. We used HAADF-STEM to understand the effect of the hydrosilylation and hydrolysis reaction times and the molecular concentration in solution on surface coverage. We used reaction times from 15 min to 24 hours for the hydrosilylation of Re-^Bubpy and Re-Phen (Molecular structure: Fig. 2a, plot: Fig. 2b, images: Fig. 2d-e, and Supplementary Fig. 4), at a fixed concentration of the molecules in the solution of 2 mM. In the case of Re-^Bubpy, the average surface coverage increases from 1.24 atoms/nm² after 1 hour to 2.14 atoms/nm² after 24 hours. However, the increase in surface coverage slows down at longer times, suggesting saturation of the surface coverage. The behavior of Re-Phen is more complicated. The average surface coverage of Re-Phen increases from 2.93 atoms/nm² to 5.93 atoms/nm² as the functionalization time increases from 15 minutes to 6 hours. Still, the surface coverage at 9 and 24 hours appears to decrease. We propose that this decrease is an artifact caused by overlapping Re atoms in the image, which prevents individual Re atoms from being recognized and counted by the CNN model. In fact, at longer reaction times, Re-Phen forms multilayers as we see different Re atoms when we change the height above the Si substrate on which we focus (Supplementary Fig. 5). The faster attachment and higher surface coverage of Re-Phen compared to Re-^Bubpy, along with the observation of greater than monolayer coverage, is potentially related to the hydrosilylation mechanism. For both Re complexes, hydrosilylation of the surface likely proceeds via carbon radicals²⁷, and this radical is stabilized for Re-Phen. The easier

formation of the stabilized radical and its longer lifetime allows more Re-Phen to react with the surface as the concentration of the reactive species is higher. In some cases, multilayers of Re-Phen are formed implying that Re-Phen can react with itself. The concentration of the Re complex in the hydrosilylation reaction also influences the surface coverage. Reducing the concentration of Re-^Bubpy by half, with a fixed functionalization time of 24 hours, decreased the surface coverage of Re-^Bubpy by approximately half (Supplementary Fig. 6). When the hydrolysis reaction to attach Pt-Porph was performed for 1 hour, 4 hours, 16 hours, or 24 hours, similar averaged surface coverages of 0.21 atoms/nm², 0.19 atoms/nm², 0.18 atoms/nm², and 0.19 atoms/nm², respectively, were observed (Supplementary Fig. 7). This suggests that the surface coverage of Pt-Porph has already reached saturation in 1 hour. Notably, in all the experiments, comparing samples that were functionalized on the same day was important. Although the general trends in surface coverage were reproducible, there were slight differences in absolute coverage on different days, which is likely related to contaminants on the surface (Supplementary Fig. 8).

To corroborate the surface coverages calculated using HAADF-STEM, two other methods, cyclic voltammetry, and ICP-MS, were utilized. To ensure accuracy in the comparison, two Si wafers (100, degenerate n-type, 380 mm thickness), which were used for cyclic voltammetry and ICP-MS, respectively, and the Si grid for HAADF-STEM imaging, were functionalized with the Re complex in the same solution. Integration of the cathodic feature in the cyclic voltammogram of immobilized Re-^Bubpy (the cathodic feature was used because the cyclic voltammograms are not reversible) yielded surface coverages of 2.8 atoms/nm² for the 3-hour sample, 1.1 atoms/nm² for the 6-hour sample, and 2.7 atoms/nm² for the 24-hour sample (Supplementary Fig. 9). These values are relatively close to those measured by microscopy, although they do not show any trends with functionalization time. After dilution, the other Si wafer was digested in nitric acid to perform ICP-MS. This method showed a similar trend to the surface coverages calculated from the HAADF-STEM images. Still, the values were about 30-50% less than those determined using microscopy (Table S4), possibly due to incomplete digestion. The surface coverage directly calculated using HAADF-STEM highlights the saturation of molecular catalysts better than the other methods and is likely more accurate, especially for monolayer surface coverages. Furthermore, HAADF-STEM is non-destructive and provides information that other methods cannot offer by showing the distribution of individual atoms.

Another crucial part of the molecular catalyst that is expected to influence its spatial distribution and surface coverage is the nature of the linker connecting the complex to a surface. To probe the effects of the linker, we prepared Re(^Rbpy)CO₃Cl complexes with different-length alkyl chains between the bpy and the alkene. After immobilization, Re(4-(8-nonen-1-yl)-4'-methyl-2,2'-bipyridine)(CO)₃Cl (Re-^Nonbpy) and Re(4-(2-ethen-1-yl)-4'-methyl-2,2'-bipyridine)(CO)₃Cl (Re-^Etbpy)³², contain two and nine carbon alkyl linkers between the bpy and the surface, respectively (Fig. 2a). We also prepared Re(4,4'-(3-buten-1-yl)₂-2,2'-bipyridine)(CO)₃Cl (Re-^Bu2bpy)³², a complex that includes two alkene groups on the bpy and, therefore, can bind to the Si through two Si–C bonds. Under the same functionalization conditions, Re-^Etbpy gives higher surface coverage than Re-^Nonbpy (plot: Fig. 2c and images: Fig. 2g-h). We hypothesize that this could be either because Re-^Etbpy forms stabilized radicals like Re-Phen or due to the greater steric profile of the complexes containing nonyl linkers, meaning they occupy more space on the surface. Several clusters are seen in the images of both samples, similar to the images of Re-^bubpy and Re-Phen, suggesting some decomposition to Re(0) during attachment. Re-^Bu2bpy (plot: Fig. 2c and image: Fig. 2f) has the lowest surface coverage and the least amount of clustering of any complex. The absence of clustering could be related to the fact that it binds to the surface via two Si–C bonds and is, therefore, more stable to decomposition during functionalization. These results demonstrate that small changes to the catalyst structure play a significant role in catalyst distribution. Our ability to observe these differences will make it easier to perform structure-activity studies to understand ligand effects on surface distribution and coverage.

A major concern associated with using HAADF-STEM is the release and diffusion of the metal atoms caused by the electron beam ^42,43. We repeatedly imaged the same spot on a sample to probe beam-induced clustering. Consecutive images of Re-^Bubpy with a 24-hour functionalization time indeed showed diffusion of metal atoms and general clustering (Fig. 3A, Supplementary Fig. 10, and Movie 1). The clustering was also observed in the consecutive images of Re-Phen (Supplementary Fig. 11 and Movie 2). However, the first frame appears to capture the molecule as immobilized, as the movement of atoms is slow (further discussed in Supplementary Section 2). To quantify the clustering by the electron beam calculated the K-function of every other image from the consecutive images (Fig. 3b-c). There is an increase in the K-function at around 4 Å in the sequential images of Re-^Bubpy and Re-Phen, confirming beam-induced clustering and likely reduction to Re(0).

We explored different strategies to control the clustering of the molecules due to electron beam. The first method involved immobilizing the catalyst through hydrosilylation and performing another hydrosilylation reaction with 1-hexene to convert some residual hydrogen-terminated Si to Si-hexyl groups³³. Such “backfilled” surfaces of Re-^Bubpy and Re-Phen stayed essentially the same as the images were recorded, and clustering was not observed in the K-function (Fig. 3d-e, Supplementary Fig. 12-13, and Movie 3-4). Another approach examined the effects of the different linker groups in Re-^Etbpy, Re-^Nonbpy, and Re-^Bu2bpy. The sample with the longer nonyl linker was observed to cluster slightly more after exposure to the beam than Re-^Bubpy, while Re-^Bu2bpy with potentially two attachments clustered less (Fig. 3f-g, Supplementary Fig. 14-15, and Movie 5-6). In general, less clustering was observed for the catalysts expected to be less mobile due to a more crowded surface (backfilling) or having two linkers (Re-^Bu2bpy). In comparison, the most flexible Re-^Nonbpy showed somewhat more clustering. Samples of Re-^Etbpy, which had the highest density, showed the most prominent clustering (Fig. 3h, Supplementary Fig. 16, and Movie 7). Cryogenic HAADF-STEM measurements of the same Re-^Etbpy sample showed slower clustering, while the initial level of clustering was the same as in images at room temperature (Fig. 3I, Supplementary Fig. 17, and Movie 8). Taken together, our data shows that the beam induces clustering, but it is a slow process. This strongly suggests that our initial images provide an accurate representation of the distribution of molecules on the surface. Further, the amount of clustering is heavily dependent on the structure of the complex on the surface and whether the surface is backfilled.

We demonstrate a non-destructive and accurate method to characterize molecular metal complexes immobilized on Si using a combination of HAADF-STEM and a deep learning-based atomic detection algorithm. This methodology provides information about surface coverage and distribution that is not available using other techniques. For example, in this work, we used HAADF-STEM to study how changes to the reaction conditions for functionalization or the structure of the molecular complex affect surface coverage and distribution. We show that even small changes can lead to significant differences, which will likely impact reactivity, as evidenced by the fact that the amount of beam effect varies depending on the structure and coverage of the complex. The work described here will enable future studies in which HAADF-STEM can characterize surfaces before and after catalysis and the observed changes correlated with catalytic performance. It may also enable conclusions to be reached about specific sites on a heterogeneous surface where immobilization of a molecular catalyst leads to improved performance. Overall, this work is an important step towards making HAADF-STEM a powerful method for characterizing immobilized molecular catalysts, which is general to different catalyst types and support structures.

General methods

Immobilization experiments were performed with two kinds of Si: grids and wafers. Grids (SiMPore Inc) were made of Si and SiO_2,which contained nine 100 x 100 µm 30 nm thick n-type Si(100) windows. Grids were used exclusively for HAADF-STEM measurements. Wafers (NOVA electronic materials) were single-side polished, 380±25 μm thick, arsenic doped, 0.001-0.005 Ω cm^-1, n-type Si(100). Wafers were used for all other measurements including XPS, ICP-MS, and electrochemistry.

All commercial chemicals were used as received, except where noted. Acetonitrile was purchased from Honeywell Burdick & Jackson and used as received or sparged with argon and dispensed directly into a glovebox. Mesitylene (Alfa Aesar) was degassed via three freeze-pump-thaw cycles and stored in a N₂-filled glove box over 3Å activated molecular sieves. Chlorobenzene (Sigma-Aldrich) and benzonitrile (Sigma-Aldrich) were used as received. The solution of 10% (volume/volume) 1-hexene (Alfa Aesar) in mesitylene for backfilling was prepared outside of the glovebox, degassed via three freeze-pump-thaw cycles, and stored in a N₂-filled glove box. Tetrabutylammonium hexafluorophosphate (ⁿBu₄NPF₆, Sigma-Aldrich or Acros) was recrystallized >3 times from hot ethanol before use. NH₄OH solution (30%, JT Baker) and HCl solution (36-38%) were used as received. Milli-Q water was obtained with a MilliporeSigma™ Direct-Q™ 8 Tap to Pure and Ultrapure Water Purification System.

Re-^Bubpy, Re-^Nonbpy and Re-^Bu2bpy⁴⁴ and Re-Phen⁴⁵ were prepared according to literature methods. ^Etbpy was prepared using a literature procedure⁴⁶, and Re-^Etbpy was prepared using a modified procedure⁴⁷, substituting Re(CO)₅Br for Re(CO)₅Cl. Pt-Porph was purchased from Frontier Specialty Chemical and used without further purification.

Immobilization of complexes of the type Re-^Rbpy

All Re-^Rbpy molecules were immobilized on Si wafers and grids according to the following procedure. Wafers and grids were cleaned by immersion in a solution of 5:1:1 H₂O: 30% H₂O₂: 30% NH₄OH for 20 minutes at 80 °C, rinsed with Milli-Q water, and subsequently immersed in a solution of 5:1:1 H₂O: 30% H₂O₂:HCl for 20 minutes at 80 °C, and finally rinsed again with Milli-Q water. Caution: Hot H₂O₂ solutions can be strongly oxidizing and should be handled with care. The wafers and grids were then submerged in a solution of 6:1 NH₄F:HF buffered oxide etchant (Sigma Aldrich) for 30 seconds to remove the native SiO₂ layer before being thoroughly rinsed with Milli-Q water, dried, and immediately transferred to a N₂-filled glovebox where they were stored. Caution: Buffered HF solutions such as this should be handled with protocols similar to that used for pure HF solutions, with special precautions to avoid any skin contact and ample amounts of calcium gluconate nearby ⁴⁸. A solution of Re-^Rbpy in mesitylene (1 or 2 mM) was added to a flask with Kontes valve, and then the etched wafer and TEM grid were added. The flask was sealed with a Kontes pin before being removed from the glovebox and submerged in an oil bath set to 160 °C for 1 – 24h under a static N₂ atmosphere. The flask was then removed from the oil bath and allowed to cool before being transferred back into the glovebox. The wafers and grids were removed from the flask, rinsed with acetonitrile, and then allowed to soak in acetonitrile overnight. Wafers and grids were then removed from the acetonitrile and stored under N₂ until they were further characterized.

Samples with a 1-hexene backfill were prepared using the same procedure, but upon return to the glovebox after 24h at 160 °C, the Re-^Bubpy solution was removed and replaced with 1 mL of a solution of 10% 1-hexene in mesitylene before the flask was again removed from the glovebox and placed in an oil bath for a further 24h at 160 °C. The vessel was allowed to cool before being transferred back into the glovebox, where the wafers and grids were removed and soaked in acetonitrile overnight. The samples were then kept in the glovebox until they were further characterized.

Immobilization of Re-Phen

Re-Phen molecules were immobilized on Si wafers and grids according to the following procedure. Wafers were cleaned by immersion in a solution of 6:1:1 H₂O: 30% H₂O₂: 30% NH₄OH for 10 minutes at 70 °C, rinsed with DI water and subsequently immersed in a solution of 5:1:1 H₂O: 30% H₂O₂:HCl for 10 minutes at 70 °C. Caution: Hot H₂O₂ solutions can be strongly oxidizing and should be handled with care. The grids were immersed in acetone overnight for cleaning. Cleaned wafers and grids were then submerged in a solution of 10:1 NH₄F:HF buffered oxide etchant (Transene) for 30 seconds to remove the native SiO₂ layer before being thoroughly rinsed with Milli-Q or DI water, dried, and immediately transferred to a N₂-filled glovebox where they were stored. Caution: Buffered HF solutions such as this should be handled with protocols similar to that used for pure HF solutions, with special precautions to avoid any skin contact and ample amounts of calcium gluconate nearby⁴⁸. The etched wafers and TEM grids added to a flask with a Kontes valve containing a Teflon-coated stir bar. A solution of Re-Phen in mesitylene (2mM, 2 mL) was added to the flask, which was sealed with a Kontes pin before being removed from the glovebox and submerged in an oil bath set to 160 °C for 15 min – 24h under a static N₂ atmosphere. Stirring was kept low and care was taken to avoid having the stir bar rapidly strike the TEM grid or wafer. The flask was then removed from the oil bath and allowed to cool before being transferred back into the glovebox. The wafers and grids were removed from the flask, rinsed with acetonitrile, and then allowed to soak in acetonitrile overnight. Wafers and grids were then removed from the acetonitrile, rinsed with acetonitrile, and stored under N₂ until they were further characterized.

Samples with a 1-hexene backfill were prepared using the same procedure, but upon return to the glovebox after 24h at 160 °C, the Re-Phen solution was removed, and the vessel was rinsed thoroughly with mesitylene. A 2 mL of a solution of 10% 1-hexene in mesitylene was then added to the flask before it was again removed from the glovebox and placed in an oil bath for a further 24h at 160 °C. The vessel was allowed to cool before being transferred back into the glovebox, where the wafers and grids were removed, rinsed with mesitylene and acetonitrile, and soaked in acetonitrile overnight. Wafers and grids were then rinsed with acetonitrile before being stored under N₂ until they were further characterized.

Immobilization of Pt-Porph

TiO₂-deposited Si wafers and grids were prepared using the following methods. First, the Si grids and wafers were cleaned by immersing in acetone overnight. Then, the surface of the Si grids was thinned down to 23 nm using an Oxford 80 Plus reactive ion etcher (RIE). The RIE was operated with 20 sccm CF₄ gas flow, 65 mTorr pressure, 150 W radio frequency (RF) power, and 15 seconds etching time. The TiO₂ layer was deposited on the Si wafer and the thinned down Si grids using a Cambridge Nanotech S200 ALD. Tetrakis(dimethylam-ido)titanium (TDMAT) precursor and H₂O were injected after 0.1 seconds and 0.015 seconds, respectively, with 20 seconds intervals between each injection for deposition. The chamber pressure and the temperature were set at 140 mTorr and 150 ℃, respectively.

The TiO₂-coated wafers and grids were immersed in acetone overnight and in isopropyl alcohol for 10 minutes for cleaning. After drying the solutions, they were immersed in a 0.5 mM solution of Pt-Porph in 10:1 chlorobenzene:benzonitrile solution. The vessel was heated to 90 ºC for 1h - 24h in air and left to cool for 1h. Then, the wafer and the Si grids were removed from the solution and rinsed twice with 5 mL dichloromethane and again with 10 mL DI water. After blowing the DI water away from the samples using an N₂ spray gun, the samples were dried in a vacuum oven at 80 ºC overnight.

X-ray photoelectron spectroscopy

A Kratos Supra+ system with a monochromatic Al Kα x-ray source was used to perform XPS. Samples were secured to the sample holder with a copper bar that provided electrical contact to the sample. The sample holder was electrically grounded, and a charge neutralizer was used to prevent charging. Spectra were collected at a pass energy of 80 eV for survey scans and 20 eV for high-resolution scans, and spectra were corrected to the C 1s peak at 284.6 eV. Peak fitting was performed in ESCApe (Kratos) using a Shirley baseline method.

ICP-MS

Samples were immersed in 0.5 mL or 1 mL 20% nitric acid (trace metal grade, aq) solution for 24h at room temperature in a 20 mL scintillation vial. The digestion liquid volume depended on the expected Re surface coverage of the wafer to avoid over-dilution of already low concentration samples. The digested solution was transferred to a 5 mL or 10 mL volumetric flask and the scintillation vial was rinsed with NERL ® reagent grade water (Thermo Scientific) and then transferred to the volumetric flask. Reagent grade water was added to the volumetric flask to dilute the digested solution to 2% nitric acid to generate a stock solution, which was transferred to a trace metal-free conical tube. An aliquot of the stock solution was further diluted into a 2% nitric acid (v/v) and 0.5% hydrochloric acid (v/v) solution prepared with 18.2 MΩ water and trace metal grade acids. The samples were analyzed for ¹⁸⁵Re by an Agilent 7900 ICP-MS instrument operated in helium reaction gas mode. The instrument was calibrated with a single source 10 ppm Re standard (Spex Certiprep 2A) in a 2% nitric acid/0.5% hydrochloric acid aqueous matrix. An internal standard of ¹⁹³Ir was used to correct for shifts in the analyte signal intensity during the analysis run.

Electrochemistry

Electrochemistry with functionalized wafers was performed in a N₂-filled glove box using a previously published electrochemical cell design^44,45. The back of the functionalized wafer was scratched with a diamond-tipped etching pen to make electrical contact. Cyclic voltammograms were recorded in acetonitrile with 0.25 M ⁿBu₄NPF₆ with a Pt wire or mesh counter electrode and an Ag wire pseudo-reference electrode. Ferrocene was added at the end of the experiment and a cyclic voltammogram was recorded with a glassy carbon electrode for calibration of the pseudo-reference electrode.

TEM observation

A JEOL NEOARM S/TEM was used for the HAADF-STEM imaging at 200 kV. We used a 6C probe size, and a 40 μm condenser aperture, yielding approximately 62.5 pA probe current. We used a 4 cm camera length for imaging. The convergence angle was approximately 27 mrad. The Gatan DigitalMicrograph software was used to capture and analyze spectroscopic data.

Ripley’s K function

We analyzed our images using a spatially descriptive statistical method (K-function) to quantify if the molecules were clustered or randomly arranged⁴⁰. The function counts the number of atoms closer than a distance r from each atom and sums them to estimate K (Supplementary Movie 9). However, the atom count decreases near the image edge because the distance r extends beyond the image boundary. Accordingly, we applied Ripley’s isotropic edge correction (K_iso(r)) where the atom count is scaled by the ratio of the circumference of the search circle that lies outside of the image to that which lies inside the image⁴⁹. Subsequently, to generate the function K_dev(r) we subtracted the expected K_iso(r) for complete spatial randomness from the adjusted K_iso(r) value from our images. K_dev(r) would be expected to be completely flat for a totally random distribution. Thus, either positive or negative values of K_dev(r) indicate clustering or dispersion at a certain distance, respectively. Finally, an R package, spatstat, was used for the calculation.

Acknowledgments

This material is based upon work solely supported as part of the Center for Hybrid Approaches in Solar Energy to Liquid Fuels (CHASE), an Energy Innovation Hub funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award Number DESC0021173. The STEM analysis was carried out at the Singh Center for Nanotechnology, which is supported by the NSF National Nanotechnology Coordinated Infrastructure Program under grant NNCI-2025608 and the Laboratory for Research on the Structure of Matter (MRSEC) supported by the National Science Foundation (DMR-1720530). The XPS analysis was performed at the Chapel Hill Analytical and Nanofabrication Laboratory, CHANL, a member of the North Carolina Research Triangle Nanotechnology Network, RTNN, which is supported by the National Science Foundation, Grant ECCS-2025064, as part of the National Nanotechnology Coordinated Infrastructure, NNCI. The authors thank J. Horwath for helpful discussions.

Author contributions

Conceptualization: SJ, JLD, JMM, NH, EAS

Methodology: SJ, HSN, BLH, AJP, ND, HA, XJ

Investigation: SJ, HSN, BLH, AJP, CLD, GPB, JC

Visualization: SJ

Funding acquisition: JLD, JMM, NH, EAS, FNC

Project administration: EAS

Supervision: EAS

Writing – original draft: SJ

Writing – review & editing: SJ, HSN, BLH, CLD, XJ, JLD, JMM, NH, EAS

Copéret, C., Chabanas, M., Petroff Saint-Arroman, R. & Basset, J.-M. Homogeneous and Heterogeneous Catalysis: Bridging the Gap through Surface Organometallic Chemistry. Angew. Chem. Int. Ed. 42, 156–181 (2003). https://doi.org/https://doi.org/10.1002/anie.200390072
Copéret, C. et al. Surface Organometallic and Coordination Chemistry toward Single-Site Heterogeneous Catalysts: Strategies, Methods, Structures, and Activities. Chem. Rev. 116, 323–421 (2016). https://doi.org/10.1021/acs.chemrev.5b00373
Serna, P. & Gates, B. C. Molecular Metal Catalysts on Supports: Organometallic Chemistry Meets Surface Science. Acc. Chem. Res. 47, 2612–2620 (2014). https://doi.org/10.1021/ar500170k
Stalzer, M. M., Delferro, M. & Marks, T. J. Supported Single-Site Organometallic Catalysts for the Synthesis of High-Performance Polyolefins. Catal. Lett. 145, 3–14 (2015). https://doi.org/10.1007/s10562-014-1427-x
Hübner, S., de Vries, J. G. & Farina, V. Why Does Industry Not Use Immobilized Transition Metal Complexes as Catalysts? Adv. Synth. Catal. 358, 3–25 (2016). https://doi.org/https://doi.org/10.1002/adsc.201500846
Witzke, R. J., Chapovetsky, A., Conley, M. P., Kaphan, D. M. & Delferro, M. Nontraditional Catalyst Supports in Surface Organometallic Chemistry. ACS Catal. 10, 11822–11840 (2020). https://doi.org/10.1021/acscatal.0c03350
Copéret, C. et al. Olefin Metathesis: What Have We Learned About Homogeneous and Heterogeneous Catalysts From Surface Organometallic Chemistry? Chem. Sci. 12, 3092–3115 (2021). https://doi.org/10.1039/D0SC06880B
Bullock, R. M., Das, A. K. & Appel, A. M. Surface Immobilization of Molecular Electrocatalysts for Energy Conversion. Chem. Eur. J. 23, 7626–7641 (2017). https://doi.org/https://doi.org/10.1002/chem.201605066
Jabbour, R. et al. Multiple Surface Site Three-Dimensional Structure Determination of a Supported Molecular Catalyst. J. Am. Chem. Soc. 144, 10270–10281 (2022). https://doi.org/10.1021/jacs.2c01013
Waelder, J., Vasquez, R., Liu, Y. & Maldonado, S. A Description of the Faradaic Current in Cyclic Voltammetry of Adsorbed Redox Species on Semiconductor Electrodes. J. Am. Chem. Soc. 144, 6410–6419 (2022).
Zhou, W. et al. Single Atom Microscopy. Microsc. Microanal. Microstruct. 18, 1342–1354 (2012).
Pennycook, S. J. & Nellist, P. D. Scanning transmission electron microscopy: imaging and analysis. (Springer Science & Business Media, 2011).
Wei, H. et al. FeO_x-Supported Platinum Single-Atom and Pseudo-Single-Atom Catalysts for Chemoselective Hydrogenation of Functionalized Nitroarenes. Nature Commun. 5, 5634 (2014).
Zhang, Z. et al. Electrochemical Deposition as a Universal Route for Fabricating Single-Atom Catalysts. Nature Commun. 11, 1215 (2020).
Ji, S. et al. Matching the Kinetics of Natural Enzymes with a Single-Atom Iron Nanozyme. Nature Catal. 4, 407–417 (2021).
Zhang, N. et al. Silver Single-Atom Catalyst for Efficient Electrochemical CO₂ Reduction Synthesized from Thermal Transformation and Surface Reconstruction. Angew. Chem. Int. Ed. 60, 6170–6176 (2021).
Zhao, J. et al. A Heterogeneous Iridium Single-Atom-Site Catalyst for Highly Regioselective Carbenoid O–H Bond Insertion. Nature Catal. 4, 523–531 (2021).
Zhao, Y. et al. Highly Stable Preferential Carbon Monoxide Oxidation by Dinuclear Heterogeneous Catalysts. Proc. Natl. Acad. Sci. U.S.A. 120, e2206850120 (2023). https://doi.org/doi:10.1073/pnas.2206850120
Benson, E. E., Kubiak, C. P., Sathrum, A. J. & Smieja, J. M. Electrocatalytic and Homogeneous Approaches to Conversion of CO₂ to Liquid Fuels. Chem. Soc. Rev. 38, 89–99 (2009). https://doi.org/10.1039/B804323J
Corbin, N., Zeng, J., Williams, K. & Manthiram, K. Heterogeneous Molecular Catalysts for Electrocatalytic CO₂ Reduction. Nano Res. 12, 2093–2125 (2019). https://doi.org/10.1007/s12274-019-2403-y
Zhang, S., Fan, Q., Xia, R. & Meyer, T. J. CO₂ Reduction: From Homogeneous to Heterogeneous Electrocatalysis. Acc. Chem. Res. 53, 255–264 (2020). https://doi.org/10.1021/acs.accounts.9b00496
Wu, Y., Liang, Y. & Wang, H. Heterogeneous Molecular Catalysts of Metal Phthalocyanines for Electrochemical CO₂ Reduction Reactions. Acc. Chem. Res. 54, 3149–3159 (2021). https://doi.org/10.1021/acs.accounts.1c00200
Morikawa, T., Sato, S., Sekizawa, K., Suzuki, T. M. & Arai, T. Solar-Driven CO₂ Reduction Using a Semiconductor/Molecule Hybrid Photosystem: From Photocatalysts to a Monolithic Artificial Leaf. Acc. Chem. Res. 55, 933–943 (2022). https://doi.org/10.1021/acs.accounts.1c00564
Buriak, J. M. Organometallic Chemistry on Silicon and Germanium Surfaces. Chem. Rev. 102, 1271–1308 (2002). https://doi.org/10.1021/cr000064s
Boukherroub, R. Chemical Reactivity of Hydrogen-Terminated Crystalline Silicon Surfaces. Curr. Opin. Solid State Mater. Sci. 9, 66–72 (2005). https://doi.org/https://doi.org/10.1016/j.cossms.2006.03.006
Wolf, S. D., Descoeudres, A., Holman, Z. C. & Ballif, C. High-Efficiency Silicon Heterojunction Solar Cells: A Review. Green 2, 7–24 (2012). https://doi.org/doi:10.1515/green-2011-0018
Buriak, J. M. Illuminating Silicon Surface Hydrosilylation: An Unexpected Plurality of Mechanisms. Chem. Mater. 26, 763–772 (2014). https://doi.org/10.1021/cm402120f
Sharma, S., Jain, K. K. & Sharma, A. Solar Cells: In Research and Applications - A Review. Mater. Sci. Appl. 6, 1145 (2015). https://doi.org/10.4236/msa.2015.612113
Veerbeek, J. & Huskens, J. Applications of Monolayer-Functionalized H-Terminated Silicon Surfaces: A Review. Small Methods 1, 1700072 (2017). https://doi.org/https://doi.org/10.1002/smtd.201700072
Mitchell, S. et al. Automated Image Analysis for Single-Atom Detection in Catalytic Materials by Transmission Electron Microscopy. J. Am. Chem. Soc. 144, 8018–8029 (2022).
Ni, H. et al. Quantifying Atomically Dispersed Catalysts Using Deep Learning Assisted Microscopy. Nano Lett. (2023).
Jia, X. et al. Synthesis and Surface Attachment of Molecular Re (I) Complexes Supported by Functionalized Bipyridyl Ligands. Inorg. Chem. 62, 2359–2375 (2023).
Huffman, B. L. et al. Surface Immobilization of a Re (I) Tricarbonyl Phenanthroline Complex to Si (111) through Sonochemical Hydrosilylation. ACS Appl. Mater. Interfaces 15, 984–996 (2022).
Langner, A. et al. Controlled Silicon Surface Functionalization by Alkene Hydrosilylation. J. Am. Chem. Soc. 127, 12798–12799 (2005). https://doi.org/10.1021/ja054634n
Ladomenou, K. et al. Photochemical Hydrogen Generation with Porphyrin-Based Systems. Coord. Chem. Rev. 304, 38–54 (2015).
O'Neill, J. S., Kearney, L., Brandon, M. P. & Pryce, M. T. Design Components of Porphyrin-Based Photocatalytic Hydrogen Evolution Systems: A Review. Coord. Chem. Rev. 467, 214599 (2022).
Folkesson, B. et al. ESCA Studies on the Charge Distribution in Some Dinitrogen Complexes of Rhenium, Iridium, Ruthenium, and Osmium. Acta Chem. Scand. 27, 19 (1973).
Komiyama, M., Ogino, Y., Akai, Y. & Goto, M. X-ray Photoelectron Spectroscopic Studies of Unsupported and Supported Rhenium Using Argon-Ion Bombardment. J. Chem. Soc., Faraday Trans. 2 79, 1719–1728 (1983).
Chastain, J. & King Jr, R. C. Handbook of X-ray Photoelectron Spectroscopy. Perkin-Elmer Corporation 40, 221 (1992).
Ripley, B. D. The Second-Order Analysis of Stationary Point Processes. J. Appl. Probab. 13, 255–266 (1976).
Cao, K. et al. Imaging an Unsupported Metal–Metal Bond in Dirhenium Molecules at the Atomic Scale. Sci. Adv. 6, eaay5849 (2020). https://doi.org/doi:10.1126/sciadv.aay5849
Egerton, R., Li, P. & Malac, M. Radiation damage in the TEM and SEM. Micron 35, 399–409 (2004).
Hoelzel, H. et al. Time-Resolved Imaging and Analysis of the Electron Beam-Induced Formation of an Open-Cage Metallo-Azafullerene. Nature Chem., 1–8 (2023).
Jia, X. et al. Synthesis and Surface Attachment of Molecular Re (I) Complexes Supported by Functionalized Bipyridyl Ligands. Inorganic Chemistry 62, 2359–2375 (2023).
Huffman, B. L. et al. Surface Immobilization of a Re (I) Tricarbonyl Phenanthroline Complex to Si (111) through Sonochemical Hydrosilylation. ACS Applied Materials & Interfaces 15, 984–996 (2022).
Metz, S. & Bernhard, S. Robust photocatalytic water reduction with cyclometalated Ir (III) 4-vinyl-2, 2′-bipyridine complexes. Chemical communications 46, 7551–7553 (2010).
Kamata, R., Kumagai, H., Yamazaki, Y., Sahara, G. & Ishitani, O. Photoelectrochemical CO2 reduction using a Ru (II)–Re (I) supramolecular photocatalyst connected to a vinyl polymer on a NiO electrode. ACS applied materials & interfaces 11, 5632–5641 (2018).
Harris, J., Minor, P., Chawla, N. & Singh, S. Development and Implementation of a Hydrofluoric Acid Safety Program in an Academic Institution. ACS Chemical Health & Safety 27, 183–189 (2020).
Ripley, B. D. Statistical inference for spatial processes. (Cambridge university press, 1988).

There is NO Competing Interest.

SM1.mp4
Supplementary Movie 1
SM2.mp4
Supplementary Movie 2
SM3.mp4
Supplementary Movie 3
SM4.mp4
Supplementary Movie 4
SM5.mp4
Supplementary Movie 5
SM6.mp4
Supplementary Movie 6
SM7.mp4
Supplementary Movie 7
SM8.mp4
Supplementary Movie 8
SM9.mp4
Supplementary Movie 9
SupplementaryInformation231101.docx
Supplementary Information

Download PDF

Version 1

posted

You are reading this latest preprint version

Aberration-Corrected STEM to Determine the Surface Coverage and Distribution of Immobilized Molecular Complexes

Status:

Version 1

Abstract

Figures

Introduction

Results and discussions

Conclusion

Methods

General methods

Immobilization of complexes of the type Re-^Rbpy

Immobilization of Re-Phen

Immobilization of Pt-Porph

X-ray photoelectron spectroscopy

ICP-MS

Electrochemistry

TEM observation

Ripley’s K function

Declarations

Acknowledgments

Author contributions

References

Additional Declarations

Supplementary Files

Status:

Version 1

Aberration-Corrected STEM to Determine the Surface Coverage and Distribution of Immobilized Molecular Complexes

Status:

Version 1

Abstract

Figures

Introduction

Results and discussions

Conclusion

Methods

General methods

Immobilization of complexes of the type Re-Rbpy

Immobilization of Re-Phen

Immobilization of Pt-Porph

X-ray photoelectron spectroscopy

ICP-MS

Electrochemistry

TEM observation

Ripley’s K function

Declarations

Acknowledgments

Author contributions

References

Additional Declarations

Supplementary Files

Status:

Version 1

Immobilization of complexes of the type Re-^Rbpy