Ryan C.
Nelson
and
Jeffrey T.
Miller
*
Chemical Science and Engineering Division, Argonne National Laboratory, 9700 S. Cass Ave., Argonne, IL 60439, USA. E-mail: millerjt@anl.gov
First published on 22nd December 2011
This is a short introduction of X-ray absorption spectroscopy (XAS) and its application to homogeneous transition metal compounds and catalysts. An XAS spectrum is composed of two regions, XANES and EXAFS, which provide element-specific information on formal oxidation state and local coordination environment, respectively. For molecules with similar environments, such as a common ligand set, the energy of the absorption edge can be calibrated from standards to obtain the formal oxidation state of unknown compounds. For structurally complex coordination environments, simulated EXAFS spectra obtained from XRD or DFT-modeled structures can be used to ascertain local structural information from an experimental EXAFS spectrum. With fast data acquisition at modern synchrotrons, it is also possible to follow the kinetic transformation of homogeneous compounds under realistic reaction conditions while gaining insight into the structural and electronic changes happening at the metal atom.
![]() Ryan C. Nelson | Ryan Nelson obtained his chemistry PhD in 2006 at the University of Wisconsin—Madison, working with Prof. Clark Landis on chiral phosphine ligands and their catalytic applications. Next, he went on to study supported organometallic complexes as models for heterogeneous catalysts with Prof. Susannah Scott at the University of California, Santa Barbara. He currently works as a postdoctoral researcher for Dr Marc J. A. Johnson in the homogeneous catalysis group in the Chemical Sciences and Engineering Division at Argonne National Laboratory. His current research endeavors involve homogeneous alkyne metathesis catalysts and extending Argonne's homogeneous XAS capabilities. |
![]() Jeffrey T. Miller | Jeff Miller joined the heterogeneous catalysis group in the Chemical Sciences and Engineering Division at Argonne National Laboratory (ANL) in 2008 after retiring from the BP refining and petrochemicals research department. He received his PhD in chemistry from Oregon State University in 1980. His research interests in include catalyst synthesis and kinetics of metallic and alloy nano-particles. His research team also uses X-ray absorption techniques at the Advanced Photon Source at ANL to determine the electronic and structural properties of homogenous and heterogeneous catalysts under realistic reaction conditions. |
There are many excellent books and reviews on the theory of and equations for X-ray absorption spectroscopy.21–27 As such, only a brief qualitative overview of the process is provided in order for the reader to better understand what is measured in the experiments. XAS is similar to techniques like UV-Vis and IR spectroscopy where the absorbance at any energy is given by the Beer–Lambert law. However, in XAS the dominant absorptive process is an excitation of a core electron, typically a 1s (K edge) or 2p (L3 and L2 edges) electron. The energies of these core electron transitions are element specific, and in order to measure transition metals and atoms with higher atomic number in solution, for example, X-ray energies of greater than about 3 keV are required. Although XAS measurements are possible at lower energy, these often require vacuum or low pressure cells, thus are not used for characterization under reaction conditions. For rapid data acquisition, a high photon flux is necessary in order to obtain high-quality spectra. Given these criteria, XAS measurements must be made at a synchrotron.
A typical, normalized XAS spectrum of Pd foil is shown in Fig. 1. From the onset of X-ray absorption, referred to as the absorption edge (Eo), to approximately 30 eV beyond is the X-ray Absorption Near Edge Structure (XANES). It is in this region that a core electron is photo-excited into a bound, empty valence state. Therefore, its shape and intensity is characteristic of the oxidation state, number and types of ligands and other species bonded to the excited atom. Because the position of the edge energy is influenced by an elements electronic structure, the XANES region is often used to determine the formal oxidation state of the element. In addition, the XANES spectrum is also influenced by the number and types of ligands giving a unique fingerprint of a particular species, so it can be used to determine the amounts of different compounds in a mixture through linear combination fitting techniques. Additional factors which affect the position and shape of the XANES spectra, especially at the L3 and L2 absorption edges, are beyond the scope of this introductory presentation.
![]() | ||
Fig. 1 Normalized XAS Spectrum of Pd foil, Eo = 24.350 keV. The approximate energy regions that make up the X-ray absorption near edge (XANES) and extended X-ray absorption fine structure (EXAFS) are highlighted by the arrows and text below the spectrum. |
As the energy of the incoming X-ray increases, the photo-excited electron has sufficient energy to escape the atom. The outgoing photo-electron has a wavelength, determined by its kinetic energy, and is scattered by the electrons of neighboring atoms. The back scattered electron interferes with the photo-electron to give small oscillations in the intensity beyond the absorption edge. The scattering portion of the spectrum continues to approximately 1000 eV above the edge and is called the Extended X-ray Absorption Fine Structure spectrum (EXAFS see Fig. 1 for Pd foil). Because the oscillations in absorption are related to neighboring atoms, the EXAFS region can be analyzed to give the number and type of neighboring atoms and their distance to the absorbing atom. The scattering intensity of photo-electron falls off rapidly with distance from the central atom, making EXAFS a local structural technique, generally one to two atoms from the metal. Because EXAFS is a local structural method, long range order is unnecessary; thus amorphous and solution samples are readily determined.
The high energy X-rays necessary for XAS can penetrate reactor window materials and solvents composed of light elements, e.g., C, H, N, O and F, with little attenuation in intensity. (Solvents that contain heavier, second-row atoms, such as CH2Cl2 or Me2SO, absorb a significant amount of X-ray radiation and as such are not suitable for these types of studies.) Thus, compounds can be investigated under a variety of conditions, such as in solution, under high pressure gasses, at elevated temperatures, and in the presence of reactive substrates. In addition, all absorbing atoms are detected simultaneously, providing an average spectrum for all species in the sample. Thus, in order to obtain meaningful information, uniform samples are essential, although under some conditions information about mixture compositions may be determined as will be demonstrated in the examples.
The purpose of this short perspective is to demonstrate the potential of XAS techniques for providing electronic and structural information on organometallic and homogeneous transition metal catalysts. Because many of these compounds are air and moisture sensitive, the design of a simple experimental reaction cell, which can be operated over a wide range of conditions, is given. Specific examples illustrate the application of the technique to the investigation of these compounds and catalysts at both steady-state and under dynamic conditions. The examples are chosen to demonstrate the kinds of information that can be obtained and also to highlight some of the limitations of the technique. The techniques described below represent standard methods that any chemist can learn; however, it is initially difficult to conduct these measurements and perform the analysis by one's self. It is suggested that new users, seek the advice and collaboration of experts who will assist with data acquisition and analysis.
![]() | ||
Fig. 2 Pd foil K-edge XANES spectrum (upper solid blue line), its first derivative (dashed red line), and the two-Gaussian fit of the first derivative (lower solid black line). The vertical lines indicate the energies of the two Gaussian functions in the fit. |
As an example of this type of analysis, the XANES spectra for three standard Pd reference compounds–Pd foil, PdCl2, and K2PdCl6–are shown in Fig. 3. There is a clear shift to higher energy in the absorptive edge with increasing Pd oxidation state: Pd(0) < Pd(II) < Pd(IV). In addition, there is a linear correlation between the formal oxidation state of Pd and edge energies (Fig. 3 inset). Although the absolute energy of these measurements is several thousand electron volts, edge differences of only a few electron volts are easily and reliably measured. With careful energy calibrations and modern synchrotron optics and detectors, reproducible energy measurements of approximately 0.1 eV are possible. For organometallic compounds where the ligand coordination environment is more complex, additional considerations are required in order to determine the formal oxidation state, as demonstrated in additional examples presented below.
![]() | ||
Fig. 3 Pd K-edge XANES spectra (upper) and their first derivatives (lower) for Pd foil (blue dotted, Eo = 24.3500 keV), PdCl2 (black solid, Eo = 24.3540 keV), and K2PdCl4 (red dashed, Eo = 24.3589 keV). The vertical lines indicate the positions of the edge energies from a two-Gaussian fit to the derivative spectra. Insert: the correlation of the edge energy (relative to Pd foil) of the reference compounds versus their formal oxidation state: slope = 2.22; intercept = −0.15. |
For many 4d and 5d transition elements, the XANES edge energy is readily determined as shown above for Pd. Many 3d (and some 4d) transition compounds, however, have pre-edge features, which are small peaks at lower energy than the edge. These features arise from distortions in the coordination geometry allowing for weak, dipole-forbidden electronic transitions. For peaks which do not overlap the XANES, the energy of the pre-edge peak has been used to identify the transition metal oxidation state.28–30 For pre-edge features which overlap the XANES, the edge energy can be more accurately determined by fitting the first derivative spectrum with a linear combination of Gaussian functions as described above.
![]() | ||
Fig. 4 The k2-weighted k-space EXAFS spectrum of Pd foil (A) and its corresponding Fourier-transformed R-space spectrum (B). The R-space spectrum was obtained by Fourier transform of the k region from 2–14.5 Å−1, and the solid and dotted lines are the magnitude and imaginary part of the Fourier transform, respectively. |
The contribution to the total scattering from one type of neighbor (e.g. a set of elements at a single distance) is often referred to as a scattering path. Chi(k) is the sum of all scattering paths and is generally not fit directly. Rather the data is Fourier transformed from energy space to distance, or R-space, Fig. 4B, where R is related to the distance of neighboring atoms to the absorbing element. The Fourier transform is a complex transformation, providing real and imaginary components. The imaginary part of the Fourier transform of Pd foil is shown as the dashed line in Fig. 4B. On their own, the real and imaginary components of the Fourier transform are not easily interpreted, so it is common to convert them into an absolute value, or magnitude, spectrum. For example, the magnitude spectrum of Pd foil is shown as a solid line in Fig. 4B. For light-element scattering paths, the magnitude spectra can be approximated as radial distribution functions of atoms surrounding the absorbing element, with peaks in R-space representing the scattering from atoms at different distances from the central atom. The peak positions in the magnitude R-space EXAFS spectra, however, are shorter than the actual interatomic distances and the exact distance is determined by fitting the spectra. Heavy atom scattering paths are more complex giving multiple peaks in the R-space spectrum. For example in Fig. 4B, the large peak at 2.5 Å and the two shoulders at lower R are all due to a single Pd–Pd scattering path. Qualitatively, the magnitude of the Fourier transform is larger for a greater number of neighbors and increases to higher R for longer distances. In addition, the shape is often characteristic of the neighboring atom. This qualitative inspection is often useful in understanding the general features of the coordination geometry. Quantitative information on the type, number and distance of neighboring atoms is obtained by fitting experimental EXAFS spectra with scatting paths determined theoretically (generally using a program called FEFF31,32) or obtained from experimental compounds of known structure, e.g., Pd foil, Na2PdCl4, or PdO. Finally, scattering information is proportional to the number of electrons of the scattering atom; therefore, elements that are close in atomic number (such as C, O, and N) appear identical. Thus, the identity of scattering atoms of similar atomic number cannot be determined by EXAFS unless the bond distances are sufficiently different.
A number of solution cells for XAS analysis are reported in the literature, ranging from very simple to fairly complex,15,17,33–35 and several of these cells fulfill many of the design criteria outlined above. As an example, a custom-built, general-purpose solution XAS cell used in our laboratories is shown in Fig. 5. It is composed of two major components: a small sample cell, Fig. 5A and B, and a temperature control block, Fig. 5C.
![]() | ||
Fig. 5 A completely assembled solution XAS sample cell (A), a disassembled cell (B), and its temperature-controlled mounting block (C). The ruler in the picture gives the scale in cm. |
The sample cell is a simple design fabricated from a commercially available Swagelok VCO cap and a body machined from a polymer rod that offers good chemical and thermal stability. Depending on the maximum temperature of operation suitable materials of construction are PEEK (polyether ether ketone, shown in Fig. 5), Teflon or Vespel, which can be operated up to about 100, 200 and 250 °C, respectively. The inner cavity of the cell body is approximately 1 cm in diameter, and holds a total sample volume of less than 3 mL. A small stir bar fits inside the cell, and an external magnetic stirrer provides good mixing. The outer diameter of the body can vary depending on the application and X-ray energies, but the top of the cell must be cut to match a VCO gland, which is necessary for a proper seal with the VCO cap. Because the body is composed entirely of low-Z elements (C, H, and O), there is little attenuation of incoming X-rays by the cell walls. A cell with 1.5 mm walls was sufficient for transmission data collection on 5 mM solutions of Pt complexes at the L3 edge (11.564 keV); however, at lower concentrations, or lower energy measurements, fluorescence measurements may be required but are easily obtained. For fluorescence measurements at the lowest absorption energies, the cell can be further modified by cutting a hole in the body and epoxying a Kapton window.
The cell cap consists of a VCO blind and nut. VCO fittings are simple compression-type fittings that use O-rings to achieve a tight seal. This type of fitting can be hand tightened to provide a seal that is stable under both vacuum and elevated pressures, at least to 80 psig (and likely much higher) without leaking. The VCO blind can be used as-is for samples that only require a static headspace. Pipe threading can easily be cut into the top of the blind to attach a septum for reactant addition via syringe. Alternatively, the VCO blind can be connected to a gas manifold for continuous gas flow through the liquid, addition of an internal thermocouple, etc. (Fig. 6). With this design, multiple cells can easily be manufactured, facilitating rapid sample turnaround. Due to the variety of materials of construction, wall (or window) thickness, pressure and temperature of operation, it is recommended that the cell be tested for safe operation in the lab prior to conducting experiments on the beamline.
![]() | ||
Fig. 6 A cell cap with an injection septum (lower left) and continuous gas flow manifold (center). |
Sample mounting and temperature control are achieved using the custom built temperature control unit (Fig. 5C). This unit is fabricated from aluminum and has a bored-out shaft through the top that just fits the outer dimensions of the sample cell for maximum thermal contact (Fig. 5C1). There is a small hole in the direction of the X-ray beam through the mounting block body for transmission measurements (Fig. 5C2). Perpendicular to the incoming X-rays, there is a conical cutout that is used for simultaneous fluorescence measurements (Fig. 5C3). Temperature control is achieved by passing a temperature controlled fluid through the body of the mounting block. This fluid passes through the hose barb connections on either side of the sample holder (Fig. 5C4) and travels along a three-dimensional path through the entire body of the holder providing rapid temperature control and uniform heating. This flow network is achieved by building the sample holder out of layers of aluminum with a variety of holes and groves cut into each layer to achieve the desired fluid path (Fig. 7). The layers are stacked with thin Viton gaskets to prevent fluid leakage and tightened with four set screws at corners of the block. With the proper temperature controlled fluids, temperatures from −40 to 200 °C are readily achievable.
![]() | ||
Fig. 7 A schematic view of each layer of the temperature control unit showing the holes and groves through the body for circulation of the heating or cooling fluid (shaded blue). The layers stack with the lowest numbered layer at the bottom. |
Full design diagrams and a parts list for the reactor and temperature-controlled mounting block are available upon request.
XANES spectra for 5 mM acetonitrile solutions of 1 and its electrochemical oxidation products, 1+ and 12+, are shown in Fig. 8. Complex 12+ is thermally unstable, so spectra of this sample were acquired at approximately −20 °C. An increase in absorption edge energy is observed with an increase in formal oxidation state. These data confirm the oxidation state assignments that were made by electrochemical measurements and could serve as calibrations for future oxidation state assignments of unknown materials. In addition, with these XANES spectra of the isolated compounds, it will be possible to determine the fractional composition of these species during future in situ experiments, using linear combination fitting techniques, a process that is described in the next section. Although not discussed here, the EXAFS spectrum acquired simultaneously with each XANES spectra could also provide information on structural changes in solution, such as that observed for the oxidation of 1 (Scheme 1).
![]() | ||
Fig. 8 Pd K-edge XANES spectra (upper) and their first derivatives (lower) of 1 (blue dotted, Eo = 24.3520 keV), 1+ (black solid, Eo = 24.3573 keV) and 12+ (red dashed, Eo = 24.3596 keV). The vertical lines indicate the positions of the edge energies from a two-Gaussian fit to the derivative spectra. Insert: the correlation of the edge energy (relative to Pd foil, Fig. 3) of the reference compounds versus their formal oxidation state: slope = 3.83; intercept = −5.2. |
While there is a linear correlation between the edge energy and formal oxidation state for 1 and its oxidized derivatives, there is not, in general, a linear correlation over all Pd compounds. For example, the linear regression of edge energies verses their formal oxidation state for the Pd standards (Fig. 3 inset) is different than that seen for 1 (Fig. 8 inset). This deviation occurs because the edge energy is determined by number of factors beyond the formal oxidation state. The nature of ligand coordination sphere and the occupancy of various molecular orbitals also affect the edge energy. These additional factors can cause some complications in this type of analysis. However, reliable oxidation state assignments for organometallic complexes are readily obtained if comparisons are made between well-defined, matched sets of known complexes with similar coordination environments and types of ligands.
In collaboration with Fogg research group (University of Ottawa), the kinetics of thermal decomposition of 3 was measured by both NMR and in situ XAS measurements. The thermal decomposition of a 5 mM (toluene-d8) sample of 3 as monitored by 1H NMR spectroscopy at 50 °C is shown in Fig. 9. A fit to this data using a first-order kinetic expression (Fig. 9, solid line) gives an observed rate constant (kobs) of 0.015 min−1.
![]() | ||
Fig. 9 The thermal decomposition at 50 °C of 3 in toluene-d8 (5 mM) as monitored by 1H NMR spectroscopy. The data (red circles) are normalized integrations of the methylidene (Ru = CH2) resonance at 19.4 ppm relative to the internal standard triphenylphosphine oxide. A least squares fit to the data (black line) using a first-order kinetic model provides an observed rate constant (kobs) of 0.015 min−1. |
The decomposition of 3 was also measured by time-dependent, in situ XAS under conditions identical to those of the NMR experiment (5 mM in toluene; 50 °C). A full XAS spectrum was acquired approximately every 1.5 min over the course of the reaction. Systematic changes in the normalized XANES spectra (Fig. 10A) are observed, with isosbestic points (e.g. 22.124 and 22.145 keV) indicating a conversion with no buildup of intermediates. Reaction kinetics could be determined by monitoring the absorbance value at a single energy value; however, linear combination fitting of the full XANES spectra provide much higher data quality.54–58 For example, the fraction of 3 in each XANES spectrum was determine by a linear combination fit using spectra of fresh and fully decomposed 3 (Fig. 10B). A least squares fit using a first-order kinetic expression provides an observed rate constant (kobs = 0.012 min−1) very similar to that determined by 1H NMR. This example shows that the XAS data provide similar quality results to other more commonly used methods of analysis for organometallic compound; however, XAS can also be applied to compounds where NMR characterization methods may not be applicable, such as paramagnetic compounds. These XAS spectra provide additional, complementary details about this process as well. For example, the XANES spectra provide information about electronic changes during the reaction, and analysis of EXAFS spectra gives an indication of structural changes that occur. These analyses as applied to this system are beyond the scope of this perspective and will be presented in future publications.
![]() | ||
Fig. 10 (A) Ru K-edge XANES spectra obtained during thermal decomposition of 3 at 50 °C in toluene (5 mM), and (B) the corresponding kinetic profile (blue circles) and least squares fit (solid line). The arrows in A indicate changes in the shape of the XANES spectra during the course of the reaction. The kinetic data in B were fit using a first order rate equation (solid line) to give an observed rate constant, kobs, of 0.012 min−1. |
![]() | ||
Fig. 11 A X-ray crystallographic structure of [Pd(NCMe)3]2(BF4)2 (4).61 The BF4− counter ions were removed for clarity. Pd = pink; N = blue; C = black; H = white. |
R-space EXAFS spectra of complex 4 in the solid state and in solution (5 mM in acetonitrile) are presented in Fig. 12. A close agreement between these spectra indicates no significant structural changes occur upon dissolution. The Pd coordination environment displays two well-resolved scattering features: Pd–N (∼1.6 Å) and Pd–Pd (∼2.1 Å). The features at longer distance are due to scattering from atoms beyond the first coordination shell. This simple spectral comparison would be sufficient to conclude that the structures are the same; however, fitting the two spectra provides important insight into EXAFS structural analysis.
![]() | ||
Fig. 12 The k2-weighted R-space EXAFS spectra of 4 in the solid state (red) and a 5 mM acetonitrile solution (black). The R-space spectrum was obtained by Fourier transform of the k region from 2.5–13.0 Å−1, and the solid and dotted lines are the magnitude and imaginary part of the Fourier transform, respectively. |
EXAFS fitting generally requires a reasonable initial-guess structure, e.g., number and type of scattering atom and their bond distances. In many organometallic compounds and homogeneous catalysts, there are often several ligands with slightly different bond lengths. These similar scattering paths overlap in the EXAFS spectra and are not always easily resolved. An approach to fitting these cases is to calculate a possible model structure using molecular mechanics, density functional theory62 or simply a modified crystallographic structure. These structural details are used to model an EXAFS spectrum for comparison with the experimental data. To demonstrate the approach, the crystallographic coordinates of 4 were used directly to generate a model EXAFS spectrum. The initially simulated spectrum using all the predicted scattering paths out to 3.3 Å from the Pd is shown in Fig. 13A. This simulation is a poor fit to the experimental data for the initial Pd–N features. In the crystallographic structure of 4, there are three Pd–N bonds per Pd with two average bond distances: (1) 2.16 Å for the acetonitrile ligand trans to the Pd–Pd bond and (2) 1.99 Å for the two acetonitrile ligands trans to each other. In the initial EXAFS model these paths destructively interfere resulting in a decreased intensity of the Pd–N feature relative to the experimental spectrum. A closer agreement between the fit and the data is achieved with only minor changes to the simulated model. The most significant change is a combination of the three individual Pd–N nitrogen scattering paths into a single path with a shared bond distance and a coordination number of 3. This model provides a fit to the solution state spectrum shown in Fig. 13B. The fitted Pd–N distances are 2.00 Å for the solution spectrum and 2.01 Å in the solid-state spectrum. The error in bond distance for an EXAFS fit is normally about 0.02 Å, thus these fitted Pd–N distances are in close agreement with the average crystallographically determined Pd–N distance (2.04 Å). The discrepancies in the fit and the data at longer distances (R > 2.3 Å) are likely due to increased fluxionally and disordering of the acetonitrile ligands between the single crystal sample and the structure in solution.
![]() | ||
Fig. 13 The k2-weighted R-space EXAFS spectra of a 5 mM solution of 4 in acetonitrile (black) a fit to the data (blue). The R-space spectrum was obtained by Fourier transform of the k region from 2.5–13.0 Å−1, and the solid and dotted lines are the magnitude and imaginary part of the Fourier transform, respectively. In both plots, the data was fit in k2-weighted R-space over a range of 1–3 Å. The scattering paths used for fitting were calculated with FEFF6 using the crystallographic coordinates. (A) The first 15 paths from the FEFF calculation were used for the fit: these paths are both the single and multiple scattering paths that span a range of 1.95–3.30 Å. All of the paths shared the same parameters. Fixed parameters: S2o = 1.0; ΔR = 0.0 Å; σ2 = 0.003. Fitted parameters: ΔEo = 5.8(2.1) eV. (B) The first three Pd–N single scattering paths from the original FEFF calculation were combined into a single path with a fixed degeneracy of 3 and a fitted bond distance and Debye–Waller factor. The Pd–Pd scattering path was also given an independent Debye–Waller factor. The remaining paths from A were not modified. Fixed parameters: S2o = 1.0; ΔR = 0.0 Å; σ2 = 0.003. Fitted parameters: ΔEo = 4.9(1.3) eV; RN = 2.00 Å; σ2N = 0.0038(9); σ2Pd = 0.0046(9). |
This example demonstrates the potential and limitations of EXAFS structural analysis. While the structure of 4 was refined using the XRD crystal structure, most often one wants to determine the coordination environment for an unknown structure. In these cases, DFT provides sufficiently accurate structural details for creating a model for comparison with experimental data. While XRD will always give the most accurate and complete structural analysis, high quality single crystals suitable for refinement cannot always be obtained. EXAFS structural determination can provide useful structural details about coordination environments in amorphous solids, in solution, and on supports. By combining density function theory structural modeling and EXAFS simulations for comparison to experimental data, EXAFS may provide some structural insights when other methods are not possible. As previously stated, XAS is an average technique, so distinguishing atomic distances, especially where the atomic number is similar (e.g. C, N, and O) can be difficult. The above example underscores this drawback. 4 has acetonitrile ligands at two different bond distances, but the best EXAFS fit provides only the correct number of Pd–N bonds and an accurate average bond distance. In general, as the coordination environment become increasingly complex, fitting of the EXAFS data can become increasingly difficult.
This journal is © The Royal Society of Chemistry 2012 |