Introduction. Motivation

Transcription

1 The important thing in science is not so much to obtain new facts as to discover new ways of thinking about them. - William Lawrence Bragg 1 Introduction Motivation Biological systems have evolved the capacity to conduct a tremendous variety of chemical reactions, a small sampling of which includes C-H functionalization by the cytochrome P450s, 1 dinitrogen reduction by nitrogenases, 2 and water splitting by photosystem II. 3 In addition to being capable of effecting these reactions at ambient temperatures and pressure, biological systems are also generally highly substrate selective and able to target a reaction to a specific location on a substrate in the presence of more reactive groups. Many of these fundamental reactions are also of industrial interest. However, owing in part to their thermodynamic and/or kinetic barriers, replicating these reactivities synthetically has met with only limited success. Thus, understanding how Nature has achieved these reactivities is of fundamental importance to both biology as well as chemistry, where this knowledge could be used to design more effective catalysts. Many of the most interesting enzyme-mediated reactions occur at metal ions that have been coordinated by proteins indeed, half of all proteins are estimated to require metals for their structure and/or function. 4 In this manner, Nature can exploit the unique chemistry offered by metals, such as their propensity for one electron redox processes, to catalyze reactions. Beyond harnessing these inherent reaction propensities, the protein scaffold can further fine tune the geometric and electronic structures of the metal ions to engender very specific reactivities and the high specificities required for biological reactions. In order to mimic the reactions of these enzymes, detailed knowledge of the metal sites is critical. Learning about metalloenzyme active sites and their reaction cycles presents many experimental challenges. Intermediates are by their very nature unstable, which often precludes their accumulation to high concentrations and can require measurement at low temperatures. Furthermore, conversion from resting enzyme to active intermediate generally does not progress 100%, leaving a mixture of species in solution. Even in cases where intermediates can be generated in high purity at high concentration, the metal center remains a small site within a large protein matrix and can be difficult to selectively probe. These challenges are common throughout bioinorganic chemistry and often impose limitations on how these systems may be studied, justifying why some questions remain unanswered despite decades of investigation. Despite these difficulties, a wide range of chemical and physical techniques do exist that are capable of probing these complex systems. Perhaps the most widely used of these methods for structural characterization is x-ray crystallography, which has yielded invaluable information on countless metalloprotein systems ever since the structure of myoglobin was first solved in By their very nature as catalysts, however, enzymes are not static 1

2 Figure 1: Plot of the 1s binding energy as a function of atomic number. 9 structures and in the vast majority of cases their intermediates cannot be crystallized, eliminating crystallography as an applicable technique. The use of spectroscopy is thus necessitated for learning about these systems. In particular, spectroscopies using hard x-rays are powerful probes of bioinorganic systems. Hard x-rays, or those with energies of 5 kev, possess enough energy to penetrate into samples and excite the core electrons of transition metal atoms. The energy differences of these core level electrons between metals are sufficiently great that core-level x-ray spectroscopies are inherently element selective (Figure 1) and able to probe a metal center without interference from the surrounding protein matrix. Furthermore, these techniques are not limited to crystalline samples and can be applied to a wide range of sample environments, including solids, solutions, and gases, making them amenable for biological applications. 6-8 Because every element absorbs and emits x-rays at unique energies, even the biologically critical spectroscopically quiet metals (e.g. Cu(I), Zn(II), Na(I), Ca(II)), may be investigated with these techniques. 6 Two of the most prominent x-ray methods are x-ray absorption spectroscopy (XAS), particularly from the metal K-edge, and x-ray emission spectroscopy (XES), which will now be discussed in more detail. X-ray Absorption Spectroscopy The use of x-rays to probe chemical samples is by no means a new idea. XAS edges were, 10, 11 in fact, measured more than a century ago. The information content of these spectra, however, has taken many decades to develop. Being an absorption technique, K-edge XAS probes transitions between the metal 1s electron and the unoccupied or partially occupied molecular orbitals (MOs) (Figure 2). For first row transition metal complexes, the first available orbitals are in the metal 3d manifold; these 1s to 3d transitions are denoted as the pre-edge. The generally low intensity of these features stems from the fact that XAS intensity is governed by the dipole selection rule ( ), stipulating that the acceptor orbital must possess some amount of metal np character, with s to d transitions ( ) being formally forbidden. These features do possess intensity, however, with the clearest origin being via an electric quadrupole mechanism (selection rule of ). 12 2

3 Figure 2: A metal K-edge XAS spectrum with an MO diagram at right depicting the electronic origins of these transitions. Quadrupole selection formally allows s to d transitions, albeit at intensities that are ~100x weaker than their dipole-allowed counterparts, explaining why these features often have such low intensities. Additionally, in point groups lacking centrosymmetry, the 3d orbitals are allowed to mix directly with np orbitals, imparting a measure of dipole allowedness to these transitions. Hence, compounds in tetrahedral symmetry have significantly more intense pre-edges than their octahedral congeners. 13 In addition to information about the molecular symmetry, pre-edges also carry information about the relative metal oxidation state, with more oxidized metals giving pre-edges at higher energies. 13 If provided with enough energy to excite above the pre-edge, the excited electron may be promoted to the metal 4p or higher-lying virtual states. As these transitions are now s to p, they result in the intense absorption known as the edge. The edge region consists of a complex interplay of contributions from site symmetry, ionization potential, and local geometry, making a full quantitative interpretation of this region theoretically challenging. For the purposes here, the position of the edge can be used mainly as a gauge of oxidation state between compounds of similar ligation. In a manner similar to that of the pre-edge, as a compound becomes more oxidized, the electrons in the 1s orbital become more stabilized, needing more energy to be promoted to the 4p orbitals. This is seen experimentally as a shift of the edge to higher energy and is diagnostic for metal-centered oxidation / reduction At even higher incident energies, the 1s electron is completely ionized from the metal and is perhaps better viewed as a photoelectron wave. This photoelectron wave can scatter off the electrons of nearby atoms and can give rise to constructive and destructive interference, seen as oscillations in the high energy region of the XAS spectrum that are termed extended x-ray absorption fine structure (EXAFS). The EXAFS contain information about the number and identity of atoms bound to the metal center, making it a valuable tool for structural analysis in non-crystalline (and particularly biological) systems. Despite being a useful probe of local structure, especially in enzyme systems, EXAFS will not be further examined in this thesis. X-ray Emission Spectroscopy Following a K-edge absorption event, the metal is left with a hole in the 1s orbital. As a very high energy configuration, this state rapidly decays with an electron from a higher level dropping to fill the 1s hole. This transition can occur either radiatively or nonradiatively with 3

4 Figure 3: An XES spectrum showing the Kα transitions along with expanded views of the Kβ and valence-to-core regions. The MO diagram at right shows the electronic origins for each of these features. the former being x-ray emission. Indeed, complete ionization of a 1s electron is exactly how an XES experiment begins. Similar to XAS, XES has been known as an analytical technique for over a century. The characteristic emission wavelengths for first row transition metal elements were determined already in , 20 with higher resolution measurements 21 and valence-to-core spectra 22 reported afterward. Despite this long history, however, the information content of XES spectra has developed slowly and only recently has this technique begun to find widespread chemical applications. As in XAS, the XES process adheres to the dipole selection rule, dictating that the decaying electrons must originate in donor orbitals possessing some amount of metal np character (Figure 3). Having the best overlap with the metal 1s, the metal 2p orbitals are the most likely origin for this electron and, indeed, 2p to 1s (Kα) fluorescence is the dominant signal observed. The 2p orbitals are, however, energetically and spatially isolated from the valence orbitals, so, despite the large signal, these transitions have little sensitivity to chemical changes at the metal (e.g. oxidation state, identity and number of ligands, etc) beyond a modest sensitivity to spin state. 23 To higher energy, transitions from the metal-based 3p orbitals the Kβ mainline composed of the Kβ and Kβ 1,3 features are also possible, though with an intensity ~10% that of the Kα lines. This lower signal is accompanied by a gain in chemical information, for the 3p orbitals are high enough in energy to interact with the valence even if they are not generally thought of as directly involved in bonding. The most significant of these interactions is the exchange interaction between the unpaired 3d electrons and the 3p hole present in the final state, giving rise to both high (all unpaired spins parallel) and low (3p antiparallel to 3d) spin combinations. The magnitude of the exchange interaction is dependent on the number of unpaired 3d electrons, resulting in the well-known sensitivity of Kβ mainlines to metal spin state As a result, Kβ mainlines have been used as fingerprints for metal spin state The dependence on the exchange interaction conveys more than nominal spin state, however; it also confers the mainlines with a sensitivity to the nature of the metal ligand bonds, as more covalent bonds delocalize metal d character onto the ligands to a greater extent than more ionic bonds, resulting in a greater attenuation of the 3p-3d exchange. A detailed discussion and exploration of the impact of covalency on Kβ mainlines is provided in Chapter 5, where it is established that the mainlines can be used as a 4

5 quantitative probe of metal-ligand covalency and that attempts to judge spin state from mainlines should be approached with caution. Lastly, at highest energies, is the valence-to-core (VtC) region composed of the Kβ and Kβ 2,5 transitions. In contrast to the Kα and Kβ mainlines, the peaks in this region derive from orbitals that are largely localized on the ligands rather than the metal. Intensity is conferred to VtC features via small amounts of metal np character mixing into these ligand-based orbitals, thus explaining the low intensity (~50-100x lower than the Kβ mainline) of these 26, 30 transitions. As transitions from ligand-localized MOs, the energetics of VtC peaks are dominated by ligand ionization potential and thus VtC XES can provide a map of ligand electronic structure For example, the Kβ feature is known to derive from ligand ns orbitals and, due to the large differences in ns ionization potential between atoms, this peak can be diagnostic for the identity of atoms bound to a metal center: First row atoms (C, N, O, F) can be easily identified by the Kβ feature, 31, 35 as can different protonation states of the 36, 37 same atom. In biological systems, the Kβ has been used to demonstrate the presence of oxo bridges in photosystem II 38 and a central carbide in FeMoco. 39 Beyond mere ligand identification, VtC XES is also sensitive to more subtle structural elements of metal complexes that are explored in this thesis. The dependence of VtC XES on ligand electronic structure is explored in global terms in Chapter 2 while a specific case study using the VtC region as a quantitative probe of N 2 activation in a model system is provided in Chapter 3. A previously unknown sensitivity to bond angle is revealed in a study of oxo-bridged iron dimers found in Chapter 4. Resonant X-ray Emission Spectroscopy Finally, XAS and XES can be combined into a two-dimensional experiment. Known variously as resonant x-ray emission (RXES), resonant inelastic x-ray scattering (RIXS), resonance Raman x-ray scattering (RRXS), and high energy resolution fluorescence detected (HERFD) absorption, this experiment involves the excitation of core electrons into low-lying bound orbitals, followed by detection of the subsequent x-ray emission (Figure 4). The experiment can conceptually be thought of in two ways: 1) As absorption spectra detected from specific emission features (e.g. Kα 1, Kβ ), or 2) as emission spectra obtained using resonant (e.g. 1s to 3d) excitation. The resonant nature of these spectra greatly increases the information content over their non-resonant counterparts. 40 An early example of this was provided by Hämäläinen where dramatic sharpening of the Dy L 3 -edge was seen by detecting from the Lα 1 emission feature. 41 After this demonstration, applications of this technique expanded, particularly for solid state systems, to include the collection of spin-selective XAS spectra and site selective EXAFS 45 using Kβ mainline detection. Despite the rich information content theoretically available in these spectra particularly when detecting from the valence XES features a lack of systematic studies on molecular model compounds has limited the applications of this technique. It has been suggested, for example, that ligand selective XAS data could be possible via detecting from VtC XES features (VtC-detected HERFD or 1sVtC RXES), 46 but, despite hints of what might be possible with this technique having been seen, no systematic study has been done. In Chapter 6, such a model study is undertaken using VtC-detected HERFD and reveals that dramatic modulation of the pre-edge region can be effected by detecting from different VtC XES features. The large changes which are observed suggest that VtC-detected XAS may be one possible route to obtaining ligand selective XAS spectra. Furthermore, the linesharpening available with Kα HERFD XAS is used to obtain high quality XANES spectra of soluble methane monooxygenase in Chapter 7, laying the foundation for future studies on the elusive intermediate Q. 5

6 Figure 4: An example of a two-dimensional resonant x-ray emission plane. Taking a horizontal slice through the plane gives a HERFD XAS spectrum and a vertical slice gives a resonant XES spectrum. Experimental Setup Much of the information content of XAS and XES spectra discussed above has developed as a result of relatively recent improvements in experimental instrumentation. Perhaps the most far reaching of these has been the advent and proliferation of the synchrotron lightsource, which provides many orders of magnitude more x-ray flux than is available with typical laboratory sources. With increased flux has come the ability to measure a greater variety of chemically interesting samples and the corresponding theoretical treatments needed to interpret this data. Indeed, most of the development of XAS has occurred in the last ~50 years and it is no coincidence that the first synchrotron XAS measurements were recorded using the Cornell synchrotron in Since then many facilities dedicated to x-ray science have emerged and this in turn has spurred much of the development of x-ray spectroscopy mentioned above. The actual collection of synchrotron-based XAS and XES data can be achieved with the instruments found in Figure 5. In the case of XAS, spectra are measured in one of two ways: Either by direct transmission through a sample or by monitoring a secondary fluorescence event. Direct transmission is conceptually quite simple and involves varying the energy of the incident x-rays while measuring the ratio of the x-ray intensity before the sample (I 0 ) to that passed through the sample (I 1 ), akin to a visible absorption measurement. Transmission measurements work well for solids and concentrated samples, but for dilute samples (such as protein solutions) the transmission signal is overwhelmed by the background intensity. For these measurements, fluorescence detection is required. The principle behind fluorescence detections is that, for every K-edge absorption event, a 1s hole is created on the metal. This 1s hole state is very unstable and is rapidly quenched by 6

7 Figure 5: Schematics of an XAS (left) and XES (right) experimental setup. The XAS schematic shows the instrumentation for both transmission and fluorescence detection. higher-lying electrons, which may emit fluorescent photons as they decay. While it is not universally true (particularly at metal L-edges), 52 at the metal K-edge the fluorescence emitted by the sample is generally proportional to the number of holes generated and hence the intensity of the absorption. Thus, one may measure an XAS spectrum (with significantly reduced background signal) by monitoring fluorescence. The detection in this setup is typically achieved with a solid state detector placed at 90 relative to the incident beam. The resolution of these detectors (~100+ ev) enables detection of either all fluorescence from the sample (total fluorescence yield, TFY) or only Kα fluorescence (partial fluorescence yield, PFY), both of which give spectra equivalent to those obtained from transmission measurements. X-ray emission is measured using a setup that is quite similar to that used for fluorescence detection of XAS with the only differences being 1) the incident energy is kept constant at a value well above the K-edge, and 2) the solid state detector is replaced by a high resolution crystal array spectrometer. The crystal array spectrometer used here focuses the emitted x- rays to the detector and takes advantage of Bragg s law ( ) by using diffraction from the crystals to select the energy of the emitted x-rays. Appropriate selection of crystals (to control the d spacing) and instrument angles (to control θ) allow for the XES features of interest to be detected. During measurement, the detected energy is varied by changing the position of the spectrometer and detector along the Rowland circles, as displayed in Figure 5, to change the value of θ. Instrumental resolutions of ~1 ev are attainable with these spectrometers. Experimentally, RXES / HERFD spectra are collected using a setup analogous to that used for standard x-ray emission. To collect resonant XES, however, the incident energy is fixed at the energy of a pre-edge transition rather than well above the edge and then the spectrometer is scanned to collect the XES spectrum. Alternatively, HERFD XAS is measured by setting the energy of the spectrometer to the emission feature of interest (the Kβ, for example) and then varying the incident energy. Using such a setup, the both HERFD XAS and also standard transmission or fluorescence XAS may be measured simultaneously, as is discussed in Chapter 6. 7

8 Computations From the brief descriptions presented above, it can be seen that XAS, XES, and RXES are complementary techniques and useful probes for studying the geometric and electronic structures of transition metal catalysts and metalloenzymes. For these methods to be deployed to their full potential, however, the information content of these spectra must be fully understood. Much effort has been devoted to developing these methods and, perhaps as expected, their evolution has been ongoing since the discovery of these techniques. In recent years, computational chemistry has emerged as a powerful tool for both the analysis and interpretation of these spectra. In particular, density functional theory (DFT) calculations have become particularly popular owing in large part to their offer of reasonable accuracy and relatively modest time scaling. Indeed, DFT has evolved to the point where calculations on large structural models of metalloprotein active sites 39, 53, 54 containing more than 200 atoms can be carried out on reasonable timescales, allowing the insights from chemical theory to be applied to bioinorganic systems. DFT calculations have recently been applied to x-ray spectra and, even within this relatively simple framework, excellent agreement with experiment has been found for calculated pre-edge XAS 55, 56 and VtC XES 31, 57 spectra. The one-electron nature of these calculations makes them ideally suited to interpret x-ray spectra in an intuitive MO picture, allowing for spectral trends to be rationalized in terms of electronic structure. Applications of DFT calculations to understanding x-ray spectra are found in every following chapter, with Chapter 2 giving a clear example of how these computations can aid in understanding spectra. Of course, limitations to these calculations do exist. For example, while this simple DFT approach has proven very effective for VtC XES, it is manifestly insufficient to describe the Kβ mainline. 31 The reason for this lies in the very nature of the DFT calculations employed: A one-electron picture is simply not able to capture the multiplet effects that dominate the mainline spectra. To do so, more complex calculations must be used, such as the restricted active space protocol developed in Chapter 5. Similar complications are encountered when attempting to calculate RXES planes, where multiplet and interference effects are both operative. For these calculations, a zoomed-out view of the plane generally appears to capture the important spectral features, including the number and energy splittings of the peaks, though close inspection reveals severe deficiencies in the intensities. A detailed comparison of these computations to experimental data and a discussion of their shortcomings are presented in Chapter 6. Clearly, quantum chemical calculations can greatly aid in the interpretation and analysis of x-ray spectra and can lend support to spectral assignments. It is this combination of experiment and theory to elucidate the chemical information of XAS and XES spectra that forms the basis for the work in this thesis. Current Work The following thesis focuses primarily on developing the information content of XES spectra and on applying XES and XAS to interesting chemical systems. Chapter 2 details the establishment of VtC XES as a general probe of ligand electronic structure using molecular orbital theory arguments and DFT calculations. Building upon this knowledge, Chapter 3 then demonstrates that VtC XES can be used to quantitatively monitor N-N bond activation in a model system. Chapter 4 explores the more subtle sensitivities of VtC spectra and shows that in addition to the well-established bond length dependence the intensities of VtC features also depend on the Fe-O-Fe bond angle for a series of iron dimers. Chapter 5 moves 8

9 to the Kβ mainline and uses a combination of experimental data and restricted active space calculations to establish the mainline as a quantitative probe of metal-ligand covalency. The techniques of XES and XAS are combined in Chapter 6, where VtC-detection is used to increase the information content of XAS spectra for a series of manganese model compounds. In Chapter 7, Kα HERFD is used to obtain high resolution Fe K-edge XAS for methane monooxygenase. Lastly, Chapter 8 examines the effects of ligand atom substitution on the electronic structures of Fe-S clusters and postulates how light atom substitution might be relevant at the FeMoco cluster of nitrogenase. 9