Polymer thin films as universal substrates for extreme ultraviolet absorption spectroscopy of molecular transition metal complexes

Polystyrene and polyvinyl chloride thin films are explored as sample supports for extreme ultraviolet (XUV) spectroscopy of molecular transition metal complexes. Thin polymer films prepared by slip-coating are flat and smooth, and transmit much more XUV light than silicon nitride windows. Analytes can be directly cast onto the polymer surface or co-deposited within it. The M-edge XANES spectra (40–90 eV) of eight archetypal transition metal complexes (M = Mn, Fe, Co, Ni) are presented to demonstrate the versatility of this method. The films are suitable for pump/probe transient absorption spectroscopy, as shown by the excited-state spectra of Fe(bpy)₃²⁺ in two different polymer supports.

1. Introduction

XANES spectroscopy is a powerful tool for measuring the electronic structure of transition metal complexes. The 3d metals are most commonly probed using synchrotron sources at the K and L_2,3 edges, corresponding to 1s→valence and 2p→valence transitions. M-edge XANES, which probes 3p→valence transitions with energies between 30 and 100 eV, is much rarer due to the short penetration depth of extreme ultraviolet (XUV) photons. Over the past few years, however, the maturation of femtosecond tabletop XUV sources has renewed interest in this spectral range. The large overlap between 3p and 3d orbitals contributes to highly feature-rich and informative spectra, from which details on the oxidation state, spin state and ligand field of the metal-containing compound can be extracted (Groot & Kotani, 2008; Zhang et al., 2016). M-edge XANES has been successfully used to measure excited-state dynamics in transition metal oxides (Vura-Weis et al., 2013; Jiang et al., 2014; Cirri et al., 2017; Schiffmann et al., 2020) and coordination complexes (Chatterley et al., 2016; Ryland et al., 2018, 2019; Ash et al., 2019; Zhang et al., 2019). XUV transient absorption spectroscopy has also been performed on semiconductors such as Si, Ge and organohalide lead perovskites (Cushing et al., 2019; Lin et al., 2017; Principi et al., 2018; Cushing et al., 2020).

Further adoption of M-edge XANES for molecular samples is hindered by the difficulty of sample preparation. In L- and K-edge absorption spectroscopy, samples are commonly prepared as a fine powder spread by hand onto layers of Kapton (polyimide) tape, or simply pressed into 1 mm sample mounts with windows of the same material (Bunker, 2010). Solution-phase spectroscopy is also possible using fluid jets or flow cells (Wilson et al., 2001; Sham et al., 1989; Smith & Saykally, 2017). Unlike soft and hard X-rays, however, the attenuation length of XUV light is of the order of only tens of nanometres (Henke et al., 1993); sample manipulation at this scale is challenging and the aforementioned preparation procedures do not easily apply. To date, XUV absorption spectroscopy has been limited to gas-phase molecules or those that can be deposited onto a suitable ultrathin substrate using gas-phase methods such as thermal evaporation.

Silicon nitride (`SiN', nominally Si₃N₄) is one such substrate, a material traditionally used for transmission windows in X-ray absorption spectroscopy (XAS) (Borja et al., 2016; Dwyer & Harb, 2017; Törmä et al., 2013). SiN is hard and inert, and can be fabricated with subnanometre surface roughness. SiN is not, however, particularly XUV transmissive – even a free-standing 100 nm membrane attenuates 60 eV light by 73% [see experimental trace, Fig. 2(d)]. At these thicknesses the substrates are extremely fragile and difficult to work with. Moreover, most solvents do not wet the SiN surface, which greatly hinders the casting of analyte films through techniques such as spincoating (Norrman et al., 2005). Even when spincoating does not fail, the resultant film may be deposited unevenly due to flexural standing wave patterns that spontaneously arise in the rotating SiN membrane (Advani, 1967). This film inhomogeneity interacts problematically with probe-beam spatial chirp and introduces artifacts into the spectrum (Lin et al., 2016). Physical vapor deposition processes sidestep the issues with deposition from solution, but are not suitable for delicate molecules. Thermal evaporation, for example, fails with compounds that are temperature-sensitive or have counterions; these decompose before subliming. This difficulty has inspired great creativity in sample preparation, such as the deposition of cobalt oxide nanoparticles from cryogenic He nanodroplets (Schiffmann et al., 2020), or the use of reflection–absorption spectroscopy from Fe^III complexes dissolved in glycerol (Lin et al., 2019).

In this work we introduce new methods of sample preparation that enable XUV absorption spectroscopy to be performed on a broad range of transition metal complexes that were previously inaccessible. We show that polymer films are an attractive alternative substrate to SiN. Compared with SiN, polymer thin films are flexible, are wetted by many solvents, and in the process described here are simple to fabricate without specialized equipment. Furthermore, the XUV transmission of the polymers explored here, polystyrene (PS) and polyvinyl ­chloride (PVC), are here measured to be, respectively, ∼60% and ∼100% greater than SiN for the same thickness.

Analytes may be cast directly onto a polymer thin-film surface or may be co-deposited from solution. The latter dispersal method embodies a `solid solution' in which analyte molecules are well separated from one another and is especially useful for compounds that show large electronic changes or fluorescence quenching upon aggregation in the solid state (Ito et al., 2018). Such solid-state solvation in polymer films has been shown to be important for the energetics and dynamics of optoelectronic materials (Delor et al., 2017).

2. Experimental details

2.1. M-edge XANES

The XUV probe is generated via high-harmonic generation (HHG) in an instrument described previously (Zhang et al., 2016). Briefly, a Ti:sapphire laser produces 4 mJ pulses of 800 nm light at a 1 kHz repetition rate with a pulse width of 35 fs FWHM. The IR laser pulses are focused into a semi-infinite gas cell filled with either neon or argon. Fig. S3(A) in the supporting information shows the XUV continuum created in the HHG process. The continuum has intensity in the range 40–90 eV, with the flux maximum at an energy that depends on the gas used. The 525 nm pump for transient experiments is produced by diverting a 0.7 mJ portion of the Ti:sapphire beam to a noncollinear optical parametric amplifier (TOPAS White). A stream of low-pressure nitro­gen gas was passed across the samples to avoid pump-induced heating.

2.2. Polymer thin-film fabrication

While polymer films may be prepared in many ways, this study focuses on film fabrication by `slip-coating' (Fig. 1, steps 1–3), a procedure in which liquid solution is drawn out from between sliding glass plates. Slip-coating – like dip-coating, doctor-blading or flow-coating (Stafford et al., 2006) – is a meniscus-guided deposition technique (Gu et al., 2018). The solution to be cast is loaded between two horizontal parallel plates, with the top plate freely supported by the liquid beneath (Fig. 1, step 1). Capillary forces constrain the liquid and cause it to coat the plates evenly, while the gap height is maintained approximately constant by the incompressible volume of liquid. The top plate is slipped off manually and frictional drag forces draw out the solution, leaving a wet film on the surface (Fig. 1, step 2). Evaporation yields the dry thin film. An iridescent coloration often develops at this stage due to thin-film interference (Fig. 1, step 3). The coated substrate is framed with adhesive tape which is then carefully peeled off the glass, taking the polymer with it and yielding the freestanding film (Fig. 1, steps 5–6). Alternatively, slow immersion of the plate into water at an angle of 45° releases the film from its glass substrate to float on the surface, from where it is easily retrieved. Photographs of these operations and the resultant films are shown in Fig. S1.

Figure 1 Schematic diagram of the slip-coating process and thin-film delamination by adhesive tape. (1) Polymer solution is applied to a glass slide substrate. (2) Polymer solution is evenly spread by slip-coating. (3) Evaporation. (4) In an optional step, analyte solution is cast upon the polymer-on-glass substrate. (5) Adhesive tape is applied to the perimeter of the polymer film; both are delaminated together from the substrate with a peeling motion. (6) This yields the freestanding thin film.

Slip-coating is simple, fast and inexpensive, and yields films of good quality for XUV transmission absorbance spectroscopy. Film thickness is controlled by adjusting the polymer solution concentration and the plate velocity during deposition (Landau & Levich, 1942; Davis et al., 2014), with accessible thicknesses ranging from a few tens of nanometres to several micrometres. The thickness of the prepared films is readily determined by fitting the visible-light interference pattern (Huibers & Shah, 1997), as detailed in the supporting information.

2.3. Suitability of polymer films

The ideal sample substrate has a large XUV transmissivity, is smooth and homogeneous over the length scales of both the XUV probe and the overall sample, and accommodates a wide variety of analytes either within its matrix or upon its surface. This section evaluates the degree to which polymer films fulfill these criteria.

2.3.1. XUV characterization

Due to the velocity gradient during manual plate separation, slip-coated polymer films show a thickness gradient over the scale of the substrate (75 mm × 25 mm glass slide), from which a region of the desired thickness may be selected (see Fig. S1). Fig. 2(a) shows the XUV absorbance profile of a typical PS film sample. The thickness varies only slightly over the 9 mm² sample area, with a standard deviation of 2.2%. On the even smaller scale of the ∼75 µm FWHM XUV beam, the film thickness is essentially constant – an important criterion for mitigating spectral artifacts which arise from sample and probe-beam spatial in­homogeneities (Lin et al., 2016). Unlike films prepared from spincoating on flexible substrates, there is no evidence of standing wave patterns in the thickness profile. While films produced in this way are homogeneous on the scales necessary for XUV spectroscopy, greater homogeneity and reproducibility could be achieved, at the cost of greater complexity, by using a computer-controlled actuator to move the plates at a precise velocity (Stafford et al., 2006).

Figure 2 (a) A 225 nm polystyrene thin film mounted on an empty Si frame. The XUV absorbance was sampled on a 200 µm interval grid to build up an image. Contour lines indicate 1% changes in relative absorbance. (b) Polymer and SiN film XUV spectra. Black lines are power law fits to the data. (c) The XUV absorbance of PS as a function of thickness. The solid line is a linear fit to the data with a y intercept of 0 and a slope of (3.61 ± 0.08) × 10^−3 nm^−1. (d) XUV spectra of 100 nm thick samples, as predicted from CXRO data (dotted lines), versus spectra constructed from the experimentally determined absorption coefficients (solid lines).

The XUV absorbance spectrum was acquired and compared with that of 100 nm SiN membranes. Fig. 2(b) shows the XUV spectrum of these materials, which in the energy range observed comprises only non-resonant absorption due to photoionization of valence electrons. This photoionization is well approximated by a power law and the absorbance is found to be directly proportional to thickness, as shown in Fig. 2(c). While the polymers absorb significantly less XUV radiation than does SiN per unit thickness (∼40% less for PS and ∼50% less for PVC at 60 eV), calculations based on atomic scattering values [shown in Fig. 2(d)] underestimate polymer absorbance and overestimate SiN absorbance (CXRO Database; Henke et al., 1993). In the case of the polymers, this discrepancy might be accounted for by a different thin-film density from that of the bulk polymer, or by an increased photoionization cross section of the polymer molecular orbitals compared with isolated atoms (Vignaud et al., 2014). In contrast to the simulation of Si₃N₄, the material used, `SiN,' is actually substoichiometric in nitro­gen. Indeed, a formulation of Si₃N_3.4 better fits the data.

2.3.2. Polymer films as substrate

PS and PVC films present a surface more easily wetted by organic solvents than SiN. Before the film is delaminated, various analyte compounds may be cast upon it by e.g. spin-, slip- or drop-casting. However, exposure to solvents that dissolve or swell the polymer (such as di­chloro­methane or tetra­hydro­furan) mars the surface and/or prevents the film from delaminating. This can be avoided by appropriate solvent choice and by reducing the time the solvent is in contact with the polymer. Typically, we accomplish this by spin-coating (Fig. 1, step 4) with immediate application of a heat gun to remove solvent quickly, which also limits the crystallite size and results in a smoother film. We found that a 4:2:1 mixture of methanol:propan-2-ol:butanol is an effective solvent system for many analytes; it also evaporates quickly and efficiently wets the polymer surfaces, yet is slow to mar them.

After a second layer is cast upon the polymer base layer, the resulting bilayered film can be delaminated in one piece. PVC films readily delaminate from glass with adhesive tape (Fig. 1, step 5). Alternatively, water flotation delaminates both PVC and PS. In some cases, PS films may also be removed with adhesive tape, though they generally adhere to the glass more strongly. All PVC bilayer films prepared in this study were removed with adhesive tape.

2.3.3. Polymer films as matrix

Samples may be prepared by co-deposition into a single-layer film, in which the polymer acts as a bulk matrix supporting the analyte. With high analyte loading, the resultant films are no longer removable with adhesive tape and are prone to tearing. Such films must be delaminated by water flotation. If water-soluble, some analyte inevitably leaches out of films when prepared in this way. Very water-sensitive analytes are better prepared on PVC films, which can be delaminated mechanically.

3. Results and discussion

The versatility of this sample preparation method is demonstrated using the eight representative coordination complexes shown in Fig. 3, representing several archetypes of molecules that can now be studied easily using M-edge XANES.

Figure 3 The chemical structures of (a) cobalt cubane [Co^III4O4], (b) (^tBuN4)Ni^IICl2, (c) Fe^IIITPPCl, (d) hemin, (e) Mn^III(acac)3, (f) Co^III(acac)3, (g) Co^IICl2 and (h) Fe^II(bpy)3Cl2.

3.1. Ground-state XUV spectroscopy

The M-edge XANES spectrum of each sample was collected and all are shown in Fig. 4. Each spectrum shows a resonant absorption edge whose position is primarily determined by element identity and oxidation state. Multiplet features on top of the edge are shaped by analyte oxidation state, spin state and coordination geometry (Zhang et al., 2016). Spectra have been baselined by subtraction of a power-law fit to the pre-edge region, corresponding to the non-resonant photoionization of substrate and ligand atoms. Following the main feature, metal 3p photoionization to the continuum contributes to the non-zero cross section and diminishes approximately as a power law thereafter.

Figure 4 XUV spectra obtained from analytes deposited on the surface of PVC, or co-deposited in PS.

Figs. 4(a1) and 4(a2). The molecular cobalt oxide cluster [Co^III₄O₄](OAc)₄(py)₄ (OAc = acetate, py = pyridine), or `cubane' [Fig. 3(a)], has garnered much interest since its isolation (Beattie et al., 1998) as a model for the cobalt-phosphate (CoPi) system of Nocera (Kanan & Nocera, 2008), as well as as a potential water oxidation catalyst in its own right (Smith et al., 2015; Nguyen et al., 2015; Ullman et al., 2014; Nguyen et al., 2017). As one of the few complexes capable of supporting a stable Co^IV center (McAlpin et al., 2011), cubane and its oxidized analogs have been the subject of prior X-ray absorption studies, including Co 1s3p (Kβ) resonant inelastic X-ray scattering (RIXS), which probes the same final state as M-edge XANES (Brodsky et al., 2017; Hadt et al., 2016). In those studies, samples were prepared as either a solid powder in a 1 mm cell with Kapton windows, or as 2 mM solutions in aceto­nitrile contained within a 3D-printed spectroelectrochemical cell.

Cubane decomposes at temperatures too low (ca 120°C) for sample preparation by thermal evaporation but is easily prepared with polymer films. We report in Figs. 4(a1) and 4(a2) the ground-state M-edge XANES spectra of cubane samples, both incorporated into PS films and deposited upon PVC. In either case, the low-spin cobalt atoms of cubane exhibit a main peak at 64.3 eV and a second at 73 eV. Except for a slight spectral broadening in the cubane/PS co-deposited sample, the position and intensity of the features are very similar between these sample preparation methods, indicating a relative insensitivity to the sample environment.

Fig. 4(b). The complex (^tBuN₄)Ni^IICl₂ [^tBuN₄ = N,N′-di-tert-butyl-2,11-di­aza­[3.3](2,6)pyridino­phane] in Fig. 3(b) serves as an example of a soluble organic nickel-containing compound, and is a convenient starting point for the formation of Ni^I and Ni^III catalysts relevant to Kumada and Negishi cross-coupling reactions (Khusnutdinova et al., 2013; Zheng et al., 2014). As shown in Fig. 4(b), the M-edge XANES spectrum of (^tBuN₄)Ni^IICl₂ has two prominent peaks at 66.2 and 68.9 eV, consistent with prior reports of Ni^II compounds with triplet ground states (Wang et al., 2013; Cirri et al., 2017).

Figs. 4(c) and 4(d). Porphyrins have been studied for light harvesting (Imahori, 2004) and phototherapy (Josefsen & Boyle, 2008), and as catalysts for diverse reactions such as oxygen or hydrogen evolution (Zhang et al., 2017). The two porphyrin compounds examined here, iron(III) tetra­phenyl porphyrin chloride (Fe^IIITPPCl) and iron(III) protoporphyrin IX chloride (hemin), are shown in Figs. 3(c) and 3(d), respectively. In a previous study from our laboratory, Fe^IIITPPCl samples were prepared by thermal evaporation and the ultrafast relaxation dynamics were investigated by transient M-edge XANES (Ryland et al., 2018), but the carb­oxy­lic acid groups on hemin preclude its sublimation at reasonable temperatures.

The low solubility of Fe^IIITPPCl limits the concentration of material, and hence the signal strength, that can be achieved in a PS matrix. Deposition upon PVC was also problematic, as di­chloro­methane (DCM) – one of the best solvents for porphyrins – is not suitable for constructing bilayers in this way. Even very short exposure to DCM mars the smooth polymer surface and prevents delamination from glass. However, it was found that a thin Fe^IIITPPCl film could be spin-coated onto glass from DCM and this neat Fe^IIITPPCl film easily delaminates and floats when slowly immersed into water. This process is like that of polymer delamination and may be generally applicable to hydro­phobic glass coatings (Khodaparast et al., 2017). The neat Fe^IIITPPCl film, which is estimated to be <100 nm by its gray-to-golden reflection, is exceedingly brittle and cannot be lifted from the water's surface without destruction. However, it can be successfully picked up upon a thin film support brought up from below. Multiple layers of sample can be built up by repeating the process as desired. This method was employed to produce the PVC-supported sample of Fig. 4(c).

Hemin, in contrast, did not require this layer-by-layer process: its carb­oxy­lic acid side groups enable its facile dissolution in alkaline solutions. Samples were prepared by spin-coating upon PVC from a mixed-alcohol solution containing a small amount of tri­ethyl­amine.

As shown in Fig. 4(c), the main peak of Fe^IIITPPCl appears at 57 eV, with a smaller pre-edge feature at 53.8 eV. The position of the main peak in hemin is identical to that in Fe^IIITPPCl, but its trailing edge diminishes more slowly and widens the feature. The hemin pre-peak [Fig. 4(d)], if it exists, is not resolved.

Figs. 4(e) and 4(f). M(acetyl­acetonate)₃ complexes are classic coordination compounds, and lend themselves well to fundamental investigations into electronic structure (Diaz-Acosta et al., 2001, 2003; Carlotto et al., 2017) and validation of spectroscopic techniques (Kubin, Kern et al., 2018; Kubin, Guo et al., 2018; Zhang et al., 2016). The low-spin Co^III(acac)₃ served as a useful reference in a study showcasing the M-edge XANES spectrum of several cobalt compounds. Samples were prepared by thermal evaporation onto SiN membranes (Zhang et al., 2016). This and the high-spin Mn^III(acac)₃ are similarly used here, with samples prepared by co-deposition from PS solution. Fig. 4(e) shows the M-edge XANES spectrum of Mn^III(acac)₃, whose main feature centers at 52.6 eV. A slight shoulder cleaves from this main peak at 49.0 eV. The spectrum of the cobalt analog is shown in Fig. 4(f) and displays a three-peaked structure, with peaks at 64.0, 67 and 74 eV. In comparison with the similarly low-spin d⁶ cobalt cubane spectra of Figs. 4(a1) and 4(a2), the peaks of Co^III(acac)₃ are at similar positions, though are slightly sharper and differ in relative intensity, probably due to the more rigidly octahedral symmetry of the acetyl­acetonate complex.

Fig. 4(g). Co^IICl₂ is a high-spin polymeric ionic compound with the Co²⁺ ions assuming octahedral geometry [Fig. 3(g)]. Aside from X-ray studies motivated by fundamental interest in its ground-state electronic configuration (Kikas et al., 1999; Wang et al., 2017), Co^IICl₂ is useful as a precursor in the production of cobalt oxide and cobalt metal thin films (Väyrynen et al., 2018). The M-edge XANES spectrum of Co^IICl₂ is shown in Fig. 4(g), displaying two large features at 61.4 and 63.8 eV, and a smaller one at 58.6 eV.

Fig. 4(h). Fe^II(bpy)₃Cl₂ [bpy = 2,2′-bi­pyridine; Fig. 3(h)] and similar Fe polypyridyl complexes are the subject of intense scrutiny due to their ultrafast intersystem crossing rates. Although ruthenium polypyridyl complexes have metal-to-ligand charge-transfer (MLCT) states with lifetimes of the order of hundreds of nanoseconds to a microsecond (Juris et al., 1988), their iron congeners relax in less than 200 fs to low-energy triplet and quintet metal-centered states (Auböck & Chergui, 2015). The former are excellent chromophores in dye-sensitized solar cells and in photoredox chemistries but the latter compounds, attractively earth-abundant and in­expensive, are inefficient due to these short lifetimes (Ardo & Meyer, 2009; Wenger, 2019; McCusker, 2019). This discrepancy has driven research into better understanding the excited-state surfaces that drive these dynamics (Miaja-Avila et al., 2016; Zhang et al., 2019; Auböck & Chergui, 2015). The Fe^II(bpy)₃Cl₂ samples were here prepared by spin-coating from solution onto PVC. The ground-state M-edge XANES spectrum [Fig. 4(h)] displays three main peaks and a shoulder at 58.4, 61.6, 67.4 and 55.0 eV, respectively.

3.2. Transient XUV spectroscopy

Efforts to improve iron(II) polypyridyl complexes have focused on altering the problematic intermediate states on the relaxation pathway through rational ligand design (Chábera et al., 2018; Wenger, 2019; McCusker, 2019). However, directly observing and characterizing these states requires a technique with ultrafast time resolution and spin sensitivity. We recently used M-edge XANES to identify an intermediate ³T state and coherent oscillations on the ⁵T_2g surface in Fe^II(phen)₃(SCN)₂ (phen = o-phenanthroline) (Zhang et al., 2019). Unlike the fortuitously sublimable phenanthroline compound, thin films of other iron(II) polypyridyl compounds, including the oft-studied prototypical spin-crossover compound Fe^II(bpy)₃Cl₂, are not so easily prepared.

The sample preparation methods developed here enabled collection of the transient M-edge XANES spectra of two Fe^II(bpy)₃²⁺ compounds: Fe^II(bpy)₃Cl₂ cast on PVC and Fe^II(bpy)₃(PF₆)₂ co-deposited in PS. Samples were pumped into the MLCT band at 525 nm and the difference spectra collected at delay times when the ⁵T_2g state is fully populated, between 1.0 and 2.0 ps (Auböck & Chergui, 2015). Fig. 5 shows the difference spectra for these two Fe^II(bpy)₃²⁺ samples compared with the previously published ⁵T_2g difference spectrum for Fe^II(phen)₃(SCN)₂. The spectra exhibit a positive excited-state absorption signal near 57.1 eV, with a shoulder at 55.3 eV that is more sharply defined in the bi­pyridine complexes. Each spectrum also displays a ground-state bleach near 67.5 eV. The successful acquisition of these spectra underscores the versatility of polymer films, as well as the aptitude of M-edge XANES spectroscopy towards the determination of excited-state electronic structure in metal complexes. The way is now made clear towards the future measurement of further spin-crossover compounds.

Figure 5 The normalized excited-state difference spectra of iron polypyridyl compounds. The red line shows Fe(bpy)3Cl2 on PVC, time-averaged between 1.0 and 2.0 ps. The blue line shows Fe^II(bpy)3(PF6)2 in PS, time-averaged between 1.0 and 2.0 ps. The black line gives the spectral component of the global fit to Fe^II(phen)3(SCN)2 data corresponding to the ^5T2g state.

4. Conclusions

The growing development of HHG sources of XUV radiation permits core-level spectroscopy to be performed using convenient in-laboratory instruments on femtosecond to attosecond timescales (Geneaux et al., 2019; Zhang et al., 2016). These advantages are balanced by the requirements that the samples be ultrathin and smooth. Limited sample preparation protocols have held back development in this area and M-edge XANES spectroscopy has so far been restricted to those materials that can be evaporated onto hard substrates such as Si₃N₄ or otherwise amenable to gas-phase experiments. We have now demonstrated a new technique for solution-processable sample deposition, enabling acquisition of M-edge spectra of many previously inaccessible compounds, shown here with the example of eight model compounds.

The polystyrene and polyvinyl chloride polymer substrates used here are easily fabricated in a simple slip-coating technique. They are flexible and highly XUV transmissive, and have more favorable wetting properties than silicon nitride. The polymers accommodate a wide variety of analytes, either directly cast upon the surface or dissolved within the bulk of the polymer thin film. Such samples are homogeneous, ultrathin and smooth, mitigating harmonic spectral artifacts and enabling the acquisition of high-quality ground- and excited-state XUV spectra.

This technique significantly broadens the types of molecular complexes that can be studied with M-edge XANES.

5. Related literature

For further literature related to the supporting information, see Andersen et al. (2001), Chakrabarty et al. (2007), Chiuzaˇian et al. (2005), Jaeger & van Dijk (1936) and Jones et al. (2013).