Opportunities and challenges of applying advanced X-ray spectroscopy to actinide and lanthanide N-donor ligand systems

Exploring the opportunities and challenges when studying actinide and lanthanide N-donor ligand systems relevant for separation technologies from the metal and ligand point of view using X-ray spectroscopy techniques and computations.

This work is focused on the n-Pr-BTP molecule and the respective An 3+ and Ln 3+ complexes with the n-Pr-BTP ligand. We applied XANES and X-ray Raman spectroscopy experimental techniques and computations with the FEFF (Rehr et al., , 2010 and FDMNES (Bunȃ u & Joly, 2009) quantum chemical codes, which are element-and bulk-sensitive probes and can be used for all Ln and An elements and N in materials in solid and liquid phase, which is not always possible with other methods. Electronic and geometric structural investigations from the 'point of view' of the metal and the ligand can facilitate understanding the fundamental principles of the N-donor ligands' selectivity. We detail experimental and computational approaches pointing out their advantages and limitations. Our aim is to guide related future studies.
Complexes with CF 3 SO À 3 (OTf) as counter ion exhibit good solubility and OTf has a lower bonding affinity to the metal than NO À 3 thereby enhancing the formation of 1:3 complexes. From time-resolved laser fluorescence spectroscopy (TRLFS) experiments Trumm et al., 2010) on Cm and Eu it can be seen that already above $ 0.6 mmol L À1 n-Pr-BTP concentration 1:3 complexes are formed exclusively. As an alternative, the [An/Ln(n-Pr-BTP) 3 ](OTf) 3 complexes were first crystallized before solving and drying on substrates. Solutions of [An/Ln(n-Pr-BTP) 3 ](OTf/ClO 4 /NO 3 ) 3 complexes have also been dried on substrates in some of the experiments. The Ln salts that have been used in the preparation of the complexes and as reference materials for the L 3 -edge measurement are hydrophilic and can contain up to nine H 2 O in the first coordination shell. Especially for experiments in solution, a high H 2 O coordination has to be assumed but could not be quantified during the measurements. Calculations have been performed with [Ln(H 2 O) 9 ](OTf) 3 . For the sake of simplicity, coordinating H 2 O molecules are not explicitly named throughout this work. Pu 3+ has been shown to be stable in [Pu(n-Pr-BTP) 3 ](NO 3 ) 3 complexes (Banik et al., 2010).

Experiments
High-resolution X-ray absorption near-edge structure (HR-XANES) spectra and core-to-core resonant inelastic X-ray scattering (CC-RIXS) maps at Ln L 3 -edges were collected at the ID26 beamline at the European Synchrotron Radiation Facility (ESRF), Grenoble, France. The X-rays emitted from the sample were energy analyzed by a Johann spectrometer in scanning geometry (Glatzel et al., 2009) and detected by an avalanche photodiode. The L 1 or L 2 emission lines of different Ln were diffracted and focused by spherically bent crystals with a 1 m radius of curvature at the corresponding Bragg angle [Eu L 1 : Ge(333)/76.75 ; Gd L 1 : Si(333)/78.28 ; Gd L 2 : Ge(620)/77.39 ].
Pu and Am L 3 -edge HR-XANES spectra have been collected at the INE beamline Walshe et al., 2014;Zimina et al., 2016Zimina et al., , 2017 at the Karlsruhe Institute of Technology (KIT) light source (KARA storage ring, KIT Campus North). The X-rays emitted from the sample were energy analyzed by a Johann spectrometer and detected by a Vortex silicon drift detector (Iwanczyk et al., 2003). The L 1 emission lines of Pu (14 282 eV) and Am (14 620 eV) were diffracted and focused by five spherically bent Si(777) crystals with a 1 m radius of curvature at Bragg angle 75.7 and 71.2 , respectively. The samples were measured in solutions of 2 mM and 8 mM An concentration, respectively. N K-edge XANES spectra were measured at the UE52-PGM beamline at the BESSY II synchrotron source of the Helmholtz-Zentrum Berlin using partial electron yield detection and a resolving power (E/ÁE) of $ 10 000. A planegrating monochromator (PGM) with a 1200 lines mm À1 grating was used to tune the incident X-rays from 380 to 420 eV. The energy of the incident X-ray beam was scanned with 0.05 eV step size. n-Pr-BTP, [Ho(n-Pr-BTP) 3 ](ClO 4 ) 3 and [Ho(n-Pr-BTP) 3 ](NO 3 ) 3 were solved in isopropanol actinide physics and chemistry 54 Tim Pruessmann et al. Actinide and lanthanide N-donor ligand systems (10 mmol L À1 ) and $ 2 mL were dried on aluminium sample holders.
Additional N K-edge XANES experiments were performed in fluorescence mode at the WERA beamline at the KIT Light Source (KARA storage ring, KIT Campus North). The incident beam was monochromatized by an SGM. The energy of the incident beam was scanned with 0.05 eV step size across the pre-edge and edge region and up to 0.5 eV in the post-edge region. The scanned energy range is from 370 to 485 eV. The spectra have been calibrated using the Ni L 3 -edge of a NiO reference. The n-Pr-BTP samples were prepared by solving the ligand in ethanol and drying $ 2 mL of the solutions on Al foil.
X-ray Raman spectroscopy experiments were performed at the 20-ID beamline at the Advanced Photon Source (APS) using the Lower Energy Resolution Inelastic X-ray scattering (LERIX) spectrometer (Fister et al., 2006). O K-edge X-ray Raman spectra of water and isopropanol and N K-edge X-ray Raman spectra of crystalline n-Pr-BTP pressed into a pellet were measured with the analyzer energy set at 9890 eV [Si(555), = 88.27 ]. The incident photon flux was 2 Â 10 12 photons s À1 using a double-crystal monochromator with Si(111) crystals. The overall experimental energy resolution was 1.3 eV estimated by measuring the FWHM of the elastically scattered radiation. The liquid samples were measured using a flow-through cell developed in cooperation with Diamond Materials GmbH (Freiburg, Germany). The cell has a diamond window with thickness of 50 mm and was set at 45 or 30 with respect to the incident beam inside a He-filled chamber. A peristaltic pump set to a flow rate of 5 mL min À1 was used to pump the liquids. For the crystalline sample, 18 detectors were employed and the sample was placed on a spinner to reduce radiation damage. Due to the cell blocking part of the scattered energy, only 15 detectors (cell at 30 ) or 12 detectors (cell at 45 ) out of 18 detectors could be used for liquid measurements.
X-ray Raman spectroscopy experiments were performed at the ID20 beamline at the ESRF using a spectrometer with 72 analyzer crystals. N K-edge X-ray Raman spectra of crystalline n-Pr-BTP and n-Pr-BTP solved in isopropanol (30 mmol L À1 ) were measured with the analyzer energy set at 9690 eV [Si(660), = 88.5 ]. The incident photon flux was $ 10 14 photons s À1 . The cell and pump setup was the same as used at the APS with the cell set at 45 to the incident beam. Due to the cell blocking part of the scattered X-rays only 24 detectors could be used for liquid measurements. The crystalline n-Pr-BTP powder was pressed into a depression in an Al plate set at 30 to the incident beam. This allowed the use of 48 crystals.

Computations
2.3.1. Structure optimization. Ln(H-BTP) 3 and An(H-BTP) 3 structures were optimized imposing the D 3 point group at the density functional theory (DFT) level with the BP86 functional (Becke, 1988;Perdew, 1986) employing the resolution-of-the-identity (RI) routines as implemented in TURBOMOLE (TURBOMOLE, 2012). Def2-TZVP basis sets (Weigend et al., 1998) were used for the light elements whereas the def-TZVP basis sets (Eichkorn et al., 1997) and small-core Stuttgart-PP were taken for the f-elements. Structures optimized in this way are denoted structure 1 throughout this work. H-BTP is used as an approximation for n-Pr-BTP as it has been proved that the substituents on the triazine rings hardly affect the complex structure (Berthet et al., 2002;Denecke et al., 2005;Petit et al., 2006).
Additionally Gd(H-BTP) 3 structures in gas and aqueous phase (called structure 2 and structure 3, respectively) have been calculated. The structures were optimized on the DFT level in C 1 symmetry using a SVP basis set and the BH-LYP functional (Becke, 1993). For the aqueous phase the structure has been solvated with COSMO-RS (Klamt, 1995).
2.3.2. Calculation of Ln/An L 3 -edge spectra. To compare different codes and methods, the spectrum of [Eu(n-Pr-BTP) 3 ](OTf) 3 was calculated with FEFF9.6 (Rehr et al., , 2010 and FDMNES (Bunȃ u & Joly, 2009). The structural data of [Eu(n-Pr-BTP) 3 ](OTf) 3 were taken from unpublished X-ray diffraction data. With FEFF, the self-consistent field (SCF) was calculated for a cluster of 28 atoms and the full multiple scattering (FMS) for a cluster of 46 atoms. A Hedin-Lundqvist-type exchange correlation potential was used. FDMNES FMS and full potential (FP) calculations were performed on a cluster of 28 atoms. In all calculations, quadrupole transitions were taken into account.
To compare different structure optimizations the experimental spectrum of [Gd(n-Pr-BTP) 3 ](OTf) 3 was compared with FDMNES FMS spectra calculated from Gd(H-BTP) 3 optimized in the different ways described above, i.e. optimization with D 3 symmetry, with C 1 symmetry and with C 1 symmetry and solvation shell.
The [Gd(n-Pr-BTP) 3 ](OTf) 3 spectrum and the f and d angular momentum projected density of states (f-, d-DOS) were calculated using the ab initio multiple-scattering theory based FEFF9.5 code (Rehr et al., , 2010. The SCF and FMS calculations were performed on a cluster of 152 atoms corresponding to one [Gd(n-Pr-BTP) 3 ](OTf) 3 molecule. A Hedin-Lundqvist-type exchange correlation potential was used. The f-DOS was calculated self-consistently by using the UNFREEZEF card. The Fermi level was set to 3 eV below the calculated value of À9.6 eV in order to reproduce the preedge structure in the spectrum. To reach convergence, the core-hole type was set to random phase approximation (RPA card) and the core-hole potential was calculated for a cluster of 47 atoms (SCREEN card).
3.1.1. Ln/An L 3 -edge HR-XANES and CC-RIXS spectra. Fig. 1 shows a simplified molecular orbital (MO) scheme of An/Ln bound to N and the excitations relevant for X-ray absorption spectroscopy (XAS) at the An/Ln L 3 -edge. At the white line (WL) electrons are excited from An/Ln 2p 3/2 states to unoccupied molecular orbitals mixtures of An/Ln 5d/6d and N 2p atomic orbitals.
The WL is mainly used to observe changes in the oxidation states (Centeno et al., 2000), i.e. the relative energy shift of the absorption edge due to the change in valence orbital occupation and the resulting change in electron density and screening of the core-hole. The WL is also sensitive to the local structure, and the density of occupied and unoccupied states, that depend also on other factors than the oxidation state, e.g. the chemical environment, i.e. type of bonding partner (Asakura et al., 2014), coordination number, local symmetry (Asakura et al., 2015), changes due to induced pressure (Rueff, 2009;Heathman et al., 2010), etc. This can lead to changes in energy position, width, shape and intensity of the WL. From a quantum chemical point of view L 3 -edge spectra describe the angular momentum projected density of states (DOS) of unoccupied d and s (less important) like states (Mott, 1949), 5d/6d states in the case of Ln/An L 3 -edge spectra following the dipole selection rules Ál = AE 1 (ÁJ = 0, AE 1). After the first experimental observation of a weak pre-edge structure in Ce XANES (Bianconi et al., 1987), band structure calculations showed a correspondence to quadrupole transitions (Finkelstein et al., 1992). Even though quadrupole transitions (2p 3/2 ! 4/5f ) are considerably weaker than dipole transitions (2p 3/2 ! 5/6d), they have been observed for a variety of Ln (2p 3/2 ! 4f transitions) and An (2p 3/2 ! 5f transitions) materials using X-ray emission spectrometers with an instrumental energy bandwidth similar to the core-hole lifetime broadening (Vitova et al., 2010(Vitova et al., , 2015(Vitova et al., , 2018Krisch et al., 1995;Kvashnina et al., 2011;Hä mä lä inen et al., 1991;Carra & Altarelli, 1990;Tanaka et al., 1994;Carra et al., 1995;Bartolomé et al., 1997Bartolomé et al., , 1999Gallet et al., 1999;Dallera et al., 2000Dallera et al., , 2003Dallera et al., , 2004Dallera et al., , 2006Journel et al., 2002;Nakazawa et al., 2002;Rueff et al., 2004Rueff et al., , 2006Glatzel et al., 2005;Sham et al., 2005;Yamaoka et al., 2006;Brouder et al., 2008;Kotani, 2008). Atomic multiplet calculations of the pre-edge spectral features of Ln L 3 HR-XANES demonstrated that they are shaped by electron-electron interactions and their shape, intensity and energy position on the excitation and emission energy depend on the number of available f electrons in the systems Severing et al., 1989;Hansmann et al., 2008). The intensity of the pre-edge can also increase in the case of hybridization of Ln f and d states.
In a standard measurement, the core-hole lifetime broadening affects the XANES spectrum leading to a significant spectral broadening compared with HR-XANES (see Fig. 2). Most notably, in HR-XANES, a pre-edge (feature A) can be resolved, the WL is sharper and has higher intensity and postedge features (B) are well separated from the WL. The HR-XANES pre-edge and WL provide details on the electronic structure not available with conventional methods. Post-edge features close to the WL are sensitive to multiple scattering of the photoelectron with the surrounding shells and therefore the local atomic geometry around the absorbing atom, e.g. bonding angles, become accessible. It can complement the information relative to interatomic distances, local structural disorder, and number and kind of atoms surrounding the absorber, which can be quantitatively obtained by analyzing the extended X-ray absorption fine-structure (EXAFS) region.
By recording HR-XANES selecting different emission lines with similar energy resolution, it is possible to obtain HR spectra with different sensitivity, due to the different intrinsic final state core-hole lifetime broadening [e.g. U L 3 HR-XANES collected at the L 5 emission line (Kvashnina et al., 2014)]. It can also be possible to detect additional spectral features due to different screening of core-holes at different energy levels. In the case of Ln L 3 -edge spectra the L 2 emission line can be used instead of the conventional highintensity L 1 emission line. However, the 4d core-hole Simplified MO scheme of An/Ln bound to N and the excitations causing the pre-edge (red), WL (blue) and EXAFS (green) features observed in An/Ln L 3 -edge spectra.

Figure 2
Standard and HR-XANES of [Eu(n-Pr-BTP) 3 ](OTf) 3 . resulting from the L 2 emission has a higher core-hole lifetime broadening (2 eV) (McGuire, 1972) than the 3d core-hole resulting from the L 1 emission (1.18 eV) (McGuire, 1974) hence no increase in resolution can be achieved. In Fig. 3 the spectra of [Gd(n-Pr-BTP) 3 ](OTf) 3 , measured recording the emitted fluorescence by fixing the outcoming energy at the maximum of the L 1 or L 2 emission lines, are compared.
The WL of the spectrum measured at the L 2 emission line is slightly narrower and the pre-edge has only one instead of two peaks. To further investigate the differences in the pre-edge region, CC-RIXS maps were recorded. Fig. 4 shows 2p3d (L 1 ) and 2p4d (L 2 ) RIXS maps of the pre-edge of [Gd(n-Pr-BTP) 3 ](OTf) 3 with a red line marking the emission energy at which the HR-XANES spectra shown in Fig. 3 were measured.
The 2p3d RIXS map has two features at the same excitation energy at $ 1180.5 eV and $ 1182.5 eV energy transfer. The HR-XANES spectrum cuts through both features leading to the observed double structure in the pre-edge. In contrast, the 2p4d RIXS map has one feature at $ 137.5 eV energy transfer and additional intensity at $ 140 eV that could be the overlap of pre-edge and WL tails or a second feature comparable with 2p3d RIXS, but with lower intensity. The HR-XANES spectrum only intersects the maximum of the first feature resulting in a single pre-edge peak. Because the differences in the CC-RIXS maps appear along the energy transfer scale, they are related to final state effects. In this case these are different splitting of the 2p 6 3d 9 5d 1 and 2p 6 4d 9 5d 1 final states, respectively; changes in the screening of the core-hole cause an energy shift of the pre-edge resonance relative to the normal emission, i.e. the maximum of the emission line in the postedge range. Since the 4f-4f electron-electron interactions have the same influence on both spectra, the appearance of two resonances along the energy transfer scale is likely related to differences in the 3d -4f and 4d -4f electron-electron interactions. It was shown that they have the highest influence on the spectrum after the 4f -4f electron-electron interactions .     The B and D absorption resonances are shifted to lower energies for the spectrum calculated with FEFF using the DFT optimized structure compared with the spectrum using the experimental structure; the former spectrum has less agreement with the experimental spectrum. This effect is probably due to the 5 pm larger Eu-N bond length in the DFT optimized structure. There is no effect on the pre-edge. For the FDMNES calculations on the other hand, the pre-edge of the spectrum using the experimental structure has larger intensity compared with the spectrum using the optimized structure, perhaps due to changes in the electronic density associated with the shorter bond length. B and C exhibit only small differences, suggesting that FDMNES treats multiple-scattering effects less accurately than FEFF. Fig. 7 shows the spectra of differently optimized Gd(H-BTP) 3 structures calculated with FEFF and FDMNES. Structures were optimized using DFT with symmetry restrictions (structure 1), without symmetry restrictions (structure 2) and without symmetry restrictions and a solvation sphere (structure 3). Gd(H-BTP) 3 structures were used instead of Eu(H-BTP) 3 structures because a solvated Eu(H-BTP) 3 structure was not available. Geometric differences between the structures are summarized in Table 1 Table 1 Angles and as shown in Fig. 8 and bond length R for differently optimized structures. Structures were optimized using DFT with symmetry restrictions (structure 1), without symmetry restrictions (structure 2) and without symmetry restrictions and a solvation sphere (structure 3).  [Gd(n-Pr-BTP) 3 ](OTf) 3 experimental spectrum compared with Gd(H-BTP) 3 spectra calculated with FEFF and FDMNES for DFT optimized structures with symmetry restrictions (structure 1), without symmetry restrictions (structure 2) and without symmetry restrictions and a solvation sphere (structure 3).

Figure 6
Calculations of [Eu(n-Pr-BTP) 3 ](OTf) 3 with FEFF and FDMNES with experimental and DFT optimized structures compared with the experimental spectrum.

Figure 8
Angles (a) and (b) describing the in-plane and out-of-plane angles of the triazine rings compared with the pyridine rings.
It can be seen that the average Gd-N bond length changes as well as the orientation of the triazine rings with respect to the pyridine ring. In the spectra calculated with FEFF a shift to higher energies of B and D can be observed from structure 1 to structure 3 showing that structure 3 is closest to the real structure. The same effect is visible in the spectra for structures 2 and 3 calculated with FDMNES. In the spectrum of structure 1, however, B and D are significantly shifted to lower energies and the pre-edge shows a double structure that is not visible in the other spectra. For the spectra compared here, the spectrum of structure 3 has the best agreement with the experimental spectrum indicating that structure 3 is closer to the real structure than the other two. Further improvements can be expected using higher-level quantum chemical methods, which are not always applicable for large molecules. Fig. 9(a) shows FEFF shell-by-shell calculations of Gd(H-BTP) 3 in a simplified approach to correlate the spectral features with specific groups of atoms surrounding the absorbing atom. The spectra of Gd(H-BTP) 3 with structure 1 were calculated with SCF and FMS radius increasing from 3 Å to 6 Å in 1 Å steps. A scheme of the used shells is shown in Fig. 9(b). The spectrum only including the contributions from the first shell, i.e. the bonding N atoms, is nearly featureless because only a few multiple-scattering paths are available. Adding the non-bonding neighbors of the bonding N atoms to the calculation as a second shell leads to clearly resolved features B and D. When adding additional atoms to the calculation, B is shifted to lower energies, farther away from the experimental energy position, and D is shifted to higher energies, closer to the experimental energy position. These shifts are due to the interference of additional scattering signals and indicate that B is mainly influenced by first-and second-shell atoms while D is related to third-and fourth-shell atoms. The pre-edge A also exhibits small changes probably due to changes in the scattering potentials of the atoms surrounding the Gd.
The origin of the pre-edge feature is revealed by calculations with the FEFF 9.5 and the FDMNES codes. The calculated Gd(H-BTP) 3 spectrum and the f-and d-DOS are plotted in Fig. 11.
The f-DOS has high intensity at the energy position of the pre-edge feature, whereas the 5d states have minor contributions. This result suggests that this feature arises from electronic transitions to orbitals with major 4f and minimal 5d Ln participations. Even though the direct bonding partners of Gd change from N in [Gd(n-Pr-BTP) 3 ](OTf) 3 to O in Gd(OTf) 3 , hardly any difference between the areas of the pre-edges actinide physics and chemistry   and WLs is detectable, indicating that the relative electronic populations of the Gd 4f and 5d states are not significantly influenced by bonding with the n-Pr-BTP molecule. Nevertheless, the À0.4 eV relative energy shift of the WL for [Gd(n-Pr-BTP) 3 ](OTf) 3 over Gd(OTf) 3 is a clear indication of better screening of the 2p core-hole due to higher electron density on Gd in [Gd(n-Pr-BTP) 3 ](OTf) 3 than in Gd(OTf) 3 . The postedge feature B is at lower energy in [Gd(n-Pr-BTP) 3 ](OTf) 3 than the corresponding feature C in Gd(OTf) 3 . In Fig. 10(a) the spectra of [Gd(n-Pr-BTP) 3 ](NO 3 ) 3 and Gd(NO 3 ) 3 are compared. The same features and energy shifts are noticeable; they are less pronounced due to the different experimental setup resulting in lower energy resolution for [Gd(n-Pr-BTP) 3 ](NO 3 ) 3 and Gd(NO 3 ) 3 . Due to the specific energy resolution of different beamlines/experimental setups and the varying intrinsic broadening of different elements it is not possible to quantitatively compare the areas of the peaks for the different compounds.

Comparison between An L 3 -edge HR-XANES of
[An(n-Pr-BTP) 3 ](NO 3 ) 3 and An(NO 3 ) 3 . [Pu/Am(n-Pr-BTP) 3 ]-(NO 3 ) 3 and Pu/Am(NO 3 ) 3 samples have been investigated using Pu/Am L 3 -edge HR-XANES. In Fig. 12 the Pu L 3 -edge HR-XANES spectra of Pu(NO 3 ) 3 and [Pu(n-Pr-BTP) 3 ](NO 3 ) 3 exhibit A and B features that are not resolved in conventional measurements. These spectral features are also characteristic of L 3 -edge HR-XANES spectra of isostructural lanthanide complexes reported previously  and also here (cf. Section 3.1.1). The A and B resonances and the WL in the Pu HR-XANES are less energy resolved due to the higher core-hole lifetime broadening contribution for An (3.3 to 4 eV) compared with Ln (0.8 to 1.6 eV). The WL is broader in the [Pu(n-Pr-BTP) 3 ](NO 3 ) 3 spectrum compared with the Pu(NO 3 ) 3 spectrum. In addition, feature B is visible only in the [Pu(n-Pr-BTP) 3 ](NO 3 ) 3 spectrum. Calculation with the FDMNES code (Bunȃ u & Joly, 2009) confirm the presence of pre-edge feature A arising from excitations to a mixture of d and f states.
In Fig. 13, features A and B are not resolved in the Am L 3edge HR-XANES Am(NO 3 ) 3 and [Am(n-Pr-BTP) 3 ](NO 3 ) 3 spectra due to the higher core-hole lifetime (Am: 3.87 eV; Pu: 3.74 eV) and experimental broadening (Am: 3.88 eV; Pu: 4.14 eV) for Am compared with Pu. The WL is broader in the [Am(n-Pr-BTP) 3 ](NO 3 ) 3 spectrum compared with Am(NO 3 ) 3 . For both Pu and Am the spectra have similar feature D 35 eV above the WL and the spectra of the complexes are shifted to lower energies (<0.5 eV). This indicates a better screening of the 2p core-hole due to higher charge density on the metal in [Pu/Am(n-Pr-BTP) 3 ](NO 3 ) 3 than in Pu/Am(NO 3 ) 3 similar to the results for the Ln compounds.

Figure 12
Pu  Fig. 14(b)] show differences in the pre-edge (A) due to the changing occupation of the 5f orbitals (5f 3 for Pu and 5f 4 for Am). Feature D in Am(H-BTP) 3 is shifted to higher energies due to the shorter An-N bond length in Am(H-BTP) 3 compared with Pu(H-BTP) 3 . Both these effects are not resolved experimentally. Fig. 15 shows N K-edge spectra of [Ho(n-Pr-BTP) 3 ](NO 3 ) 3 measured at the WERA beamline in partial electron yield (PEY), total electron yield (TEY) and fluorescence mode. PEY detection measures electrons emitted from the sample and discriminates between photo-and Auger electrons. This method is surface sensitive up to a depth of 1-2 nm. TEY detection measures all emitted electrons by counting the current necessary to neutralize the sample. Both electron detection methods are susceptible to charge build-up in nonconducting samples like n-Pr-BTP and its complexes. They are also not element selective, which is especially problematic for N K-edge spectra in the presence of C due to a large nonlinear background from the C K-edge. Thus, fluorescence detection is generally preferred for materials for N K-edge investigations in the presence of C, even though it has worse signal-to-noise ratio and possible self-absorption artifacts compared with the electron detection methods. At the UE52-PGM beamline, however, the fluorescence detector was unavailable for the beamtime, so PEY has been used. Fig. 16 shows N K-edge spectra of n-Pr-BTP measured in PEY at the UE52-PGM beamline at BESSY and at the WERA beamline at KARA. Both spectra have a similar signalto-noise ratio. The spectrum collected at the UE52-PGM beamline is shifted to higher energies compared with the spectrum collected at the WERA beamline, because at the UE52-PGM beamline no reference spectra could be collected to calibrate the energy scale. In both spectra the FWHM is $ 1 eV and spectral features are equally visible.

N K-edge XANES experiments -feasibility studies
During the measurements at BESSY II, changes of preedge features are observed after irradiation of the samples (Fig. 17). Feature A loses intensity while the intensity of feature B is increased. These effects are attributed to radiation damage, i.e. ionization and breaking of bonds by the incident X-ray beam, and were not observed at the WERA beamline (KARA) due to the lower photon flux density impinging on the sample. Fig. 18(a) shows an N K-edge spectrum of Ho(NO 3 ) 3 compared with [Ho(n-Pr-BTP) 3 ](NO 3 ) 3 . The main intensity of the Ho(NO 3 ) 3 spectrum lies at higher energies than the [Ho(n-Pr-BTP) 3 ](NO 3 ) 3 pre-edge, which is barely influenced. This is confirmed by direct comparison between NO À 3containing and NO À 3 -free {[Ho(n-Pr-BTP) 3 ](NO 3 ) 3 and [Ho(n-Pr-BTP) 3 ](ClO 4 ) 3 } complexes, see Fig. 18     N K-edge spectra of n-Pr-BTP measured in PEY at the UE52-PGM beamline at BESSY and at the WERA beamline at KARA.

X-ray Raman spectroscopy -feasibility studies
The process of separation of An from Ln takes place in a liquid phase. However, the N K-edge low photon energy ($ 400 eV) requires ultra-high-vacuum investigations challenging for liquid samples. First XAS tests with a liquid, constant flow, pump-through cell equipped with 150 nm SiC window (Blum et al., 2009) at beamline 8.0.1 at the Advanced Light Source (ALS) resulted in fast formation of X-ray induced radiolysis, i.e. radiation damage, coloring the n-Pr-BTP liquid sample from light orange to dark brown within 2 to 3 minutes. An alternative technique for K-edge XANES investigations of low-Z elements is X-ray Raman spectroscopy. An incident beam with energies above 10 keV facilitates use of liquid sample cells and double containments necessary for investigations of radioactive materials. The high energy of the X-rays reduces the radiation damage and allows penetration through windows in sample cells made from materials such as Kapton, which has high chemical and X-ray stability and therefore is often used as window material. Disadvantages of the technique are: (1) the low cross section of the process [0.13 cm 2 g À1 inelastic scattering cross section compared with 3 Â 10 4 cm 2 g À1 photo-absorption cross section (Henke et al., 1993)] requiring at least 3 at% concentration of N atoms in the sample for 2 Â 10 12 photons s À1 incident beam, 18 analyzer crystals, (2) the reduced experimental energy resolution when favoring the flux compared with a dedicated soft X-ray beamline leading to broadening of the spectral features. Only a few X-ray Raman spectroscopy measurements of highly concentrated liquid samples are reported in the literature (Bowron et al., 2000;Bergmann et al., 2002Bergmann et al., , 2007Nä slund et al., 2005;Juurinen et al., 2013Juurinen et al., , 2014Wernet et al., 2004;Pylkkanen et al., 2011;Sahle et al., 2013Sahle et al., , 2016Niskanen et al., 2015).
At the 20-ID beamline at the APS the feasibility studies carried out in November 2012 presented here were the first measurements of liquids performed with the LERIX spectrometer (Fister et al., 2006); therefore they contributed to the development program of the beamline to extend the range of possible samples. Fig. 19(a) shows the O K-edge isopropanol spectrum after averaging two scans measured 45 minutes per scan. We observed formation of gas bubbles during the measurements probably formed by radiolysis or heating of the isopropanol. The angle of the cell with respect to the incoming X-ray beam was changed from 45 to 30 to allow the bubbles to escape from the cell without disturbing the measurement. This cell arrangement reduced the number of usable analyzers from 15 to 12. In the O K-edge water spectrum [ Fig. 19(b)], details for the edge region, similar to those reported for water ice in the literature (Fister et al., 2009;Zubavichus et al., 2006), are well distinguishable. The N K-edge spectrum of crystalline n-Pr-BTP measured for 30 minutes by averaging over 18 detectors is shown in Fig. 19(c). The spectrum has a lower signal-to-noise ratio than the standard XANES spectrum (Fig. 18). After one scan (30 min) the changed surface color of the sample indicated damage by the beam. A second scan was consistently showing a lower signal. The n-Pr-BTP solution revealed no visible color changes after several hours exposure to the beam. However, due to the low concentration no discernible N signal was measured.
At the ID20 beamline at the ESRF (Huotari et al., 2017) it was possible to measure n-Pr-BTP solutions due the higher number of available analyzer crystals and the higher incident photon flux (ID20: 10 14 photons s À1 ; 20-ID: 10 12 photons s À1 ). The experiments were performed in November 2014. The effects of radiation damage in [Ho(n-Pr-BTP) 3 ](ClO 4 ) 3 N K-edge spectra after irradiation of the samples.
pre-edge and WL of the spectra have similar energy positions, but different intensities due to difficulties with the normalization. Even though the signal-to-noise ratio is much worse for the sample in solution, the general structure of the spectrum can be easily recognized and it can be assumed that the structure in solved and crystalline state does not differ significantly. The signal-to-noise ratio of the spectrum of the liquid sample can be improved by utilizing more analyzer crystals and an improved cell design, e.g. using a capillary. In addition, the solution should be temperature controlled to reduce evaporation and avoid precipitation of the complex due to temperature difference between preparation lab and beamline.

Conclusions
The Ln/An L 3 -edge HR-XANES technique reveals higher charge density on the metal Ln/An atoms for the [Ln/An(n-Pr-BTP) 3 ](OTf/NO 3 ) 3 compared with the Ln/An(OTf/NO 3 ) 3 complexes. The high energy resolution allows resolving a preedge feature for the [Ln(n-Pr-BTP) 3 ](OTf) 3 complexes and it is shown that, as indicated in previous studies (Vitova et al., 2010(Vitova et al., , 2015(Vitova et al., , 2018Finkelstein et al., 1992;Krisch et al., 1995;Kvashnina et al., 2011), it originates from electronic transitions to orbitals with predominant 4f character. The shape and energy positions of these pre-edges do not change noticeably for the [Ln(n-Pr-BTP) 3 ](OTf) 3 and Ln(OTf) 3 complexes. This strongly suggests that the 4f states are localized on the metal atom and do not participate in bonding. The shapes of the pre-edges vary and depend on the number of 4f electrons. The 4f electrons induce electron-electron interactions leading to a complex structure of the 4f states, i.e. multiplets. Due to the large core-hole lifetime broadening effects pre-edges are not resolved for the [An(n-Pr-BTP) 3 ](NO 3 ) 3 complexes. Use of the L 5 emission lines for measurements of An L 3 -edge HR-XANES spectra would lead to reduced broadening effects. The 5f states can be also directly probed by An M 4,5 -edge HR-XANES and their level of participation in the chemical bond elucidated. It is demonstrated that the HR-XANES technique allows resolving post-edge features not visible in the conventional spectra. The correlation of their energy positions to specific structural changes, i.e. interatomic distances and bonding angles, are revealed with the help of FEFF and FDMNES XANES quantum chemical calculations and simulations. Both codes are used for detailed analyses of the spectra. Optimized structures with additions of 'no symmetry restrictions' and interactions with the solvation sphere improve the agreement between theory and experiment.
The benchmark calculations using various codes and input structures helped to select the most appropriate conditions for the simulations. The input parameters for the FDMNES XANES calculations were tested and defined. We found that the best agreement between theory and experiment was achieved using structure 3 [aqueous Gd(H-BTP) 3 ], i.e. structure 3 is closer to the real structure than the other two.
The benchmark X-ray Raman spectroscopy studies demonstrate the applicability of this novel technique for investigations of liquid samples of partitioning systems at the N K-edge and K-absorption edges of other low-Z elements. No significant differences between the N K-edge spectrum for O K-edge spectra of isopropanol (a) and water (b). N K-edge spectrum of n-Pr-BTP (c).

Figure 20
N K-edge spectra of n-Pr-BTP crystalline and in isopropanol solution.
the n-Pr-BTP molecule in the solid phase or solved in isopropanol are found. This result strongly suggests that the results obtained for the solid state complexes and ligands are relevant also for their liquid forms. Specific practical suggestions for further improvement of the experimental set-up are given based on experience gained over several experiments using the state-of-the-art spectrometers installed at the brightest synchrotrons. It is shown that N K-edge X-ray Raman spectroscopy investigations of N-donor ligands in solution are in general possible at the ID20 beamline, ESRF. However, the sample setup has to be further improved by using a capillary to allow a high signal-to-noise ratio; temperature-controlled cooling of the sample will reduce the evaporation of the solvent and possible radiation damage.