1 Introduction

Solvent effects on reaction rates have a long history in catalysis [1–7]. The conventional rationalizations of solvent effects in catalysis include the modification of the solubility of reactants and products [5, 8], the modification of transport rates [8] and the effects of solvent on the adsorption of reactants [2], products [9] and the solvent itself [1]. Interest in solvent effects has increased recently [8, 10, 11] to contend with the complexity of the reaction media that arise in the catalytic processing into fuels of biogenic feedstocks, particularly those prepared from thermochemically produced biogenic feedstocks (pyrolysis oil). Pyrolysis oil usually presents initially as a blend of both polar and nonpolar compounds that transforms with increasing hydroprocessing into segregated layers of oil and water.

When a reaction medium contains components of widely differing polarizability, the composition and dynamics of a catalyst’s cybotactic region (solvent immediately adjacent to the active sites) may differ from the average composition of the bulk fluid. In the hydroprocessing of biogenic fuels, it would be useful to understand the nature of the reaction medium adjacent to the catalyst. Similarly in Fischer–Tropsch synthesis the reaction mixture may wet the catalyst differently as a function of conversion because the hydrocarbon molecules comprising the target fuel are nonpolar (relative permittivity, ε ≲ 2) while water and intermediate oxygenates water are strong to moderate dielectrics (10 ≤ ε ≤ 80). Very localized solvent gradients could also be present in reactions occurring in more conventional solvent mixtures [12–14] or ionic liquids [15] even in the absence of transport limitations that can create concentration gradients, and possibly well before the separation of immiscible reactant and product phases that occurs at high conversion. Such a thermodynamically driven segregation could arise either from preferential solvation of the sites themselves or from solvent structuring that propagates from the surface of the catalyst support. The local structuring of the solvent could affect reaction selectivities and rates by altering the energies of reaction intermediates and transition states, and would depend on interactions whose characteristic energies and dynamics lie between those of strong ligation and the solvent–solvent interactions of the bulk fluid.

Solvatochromism [16] or, more generally, perichromism [17] (effect of the surrounding medium on the Stokes shifts between absorption and emission maxima) has proved useful in biology [16] and in rheology [18] for probing compositional variations and fluid mobility at length scales of around 1 nm [19], comparable to what we envisage for stratification of multiphasic reaction media in the pores of heterogeneous catalysts [20]. UV–visible spectroscopy of surface species is well established for studying catalysts [15, 21–27] and perichromism of indicator dyes has also been used to characterize solid surfaces and catalysts [17, 19, 28–34]. However, we could find no report of the use of solvatochromism of surface-supported chromophores, or their analogs, to directly characterize the composition of the reaction medium adjacent to the sites of heterogeneous catalysts, such as might eventually be deployed in operando.

In brief, solvatochromism arises when the dipole moments of the ground state (\(\mu_{G}\)) and excited state (\(\mu_{E}\)) of a chromophore differ in magnitude and orientation such that the energies of absorption (hν A ) and fluorescence (hν F ) are differentially stabilized by the polarizability of the surrounding molecules, whose relaxations are slow on the time scale of the (vertical) absorption and emission events (Fig. 1). Although there are many refinements [16, 35–38], the effect of the solvent on the difference in energy between absorption and emission can often be correlated, through Lippert’s Relation [39], with the properties of the chromophore (changes in its dipole moment upon excitation (μ G  − μ E ), radius of the cavity (a) it makes in the solvent), and the properties of the solvent (orientation polarizability, Δf):

Fig. 1
figure 1

Schematic of vertical absorption and emission processes among the ground state, S0, the unrelaxed excited state S1′, the relaxed excited state S1 and the unrelaxed ground state S0′. The block arrows represent the dipole vectors of the molecules in the ground and excited states

Full size image
$$hv_{A} - hv_{F} = \frac{2}{c}\Delta f\frac{{\left( {\mu_{E} - \mu_{G} } \right)^{2} }}{{a^{3} }},$$
(1)

where the orientation polarizability of the environment depends on the relative permittivity of the medium, ε, and its index of refraction, n:

$$\Delta f = \frac{\varepsilon - 1}{2\varepsilon + 1} - \frac{{n^{2} - 1}}{{2n^{2} + 1}}.$$
(2)

As a first step towards determining the effects of the solid surface of a heterogeneous catalyst and its porosity on the structure and composition on the liquid phase adjacent to the catalyst sites, we measured the solvatochromism of a series of homogeneous model compounds (Fig. 2). The model complexes were charge-transfer chromophores prepared using established or analogous literature recipes to ligate oxometalates of V, Cr, Mo and W, with lacunary, polyhedral oligomeric silsesquioxane (POSS). These POSS-ligated oxometalates exhibit UV–Visible absorption spectra similar to those of the surface-supported moieties but the complexes dissolve in common solvents to form readily analyzable solutions. POSS-ligated oxometalates have proven to be close mimics of the spectroscopic and catalytic behaviors of their silica-supported analogs [40] and they are seeing increased use as model catalysts [41, 42] and as precursors to catalysts [43, 44].

Fig. 2
figure 2

Schematics of the mono-oxo (left R = phenyl) and dio-oxo POSS-ligated complexes (right R = isobutyl); M = Cr, Mo, or W

Full size image

We report here the construction and interpretation of prototypic calibration curves that assay the composition of a liquid medium near a moiety that resembles the presumed active site in silica-supported oxide catalysts, which are well known as catalysts for polymerization [45], selective oxidation [46], metathesis [47], and photolysis [48].

2 Materials and Methods

2.1 Materials

Oxo anions of vanadium, molybdenum and tungsten were prepared by the reaction of equimolar oxychlorides with trihydroxyphenyl T7 polysilsesquioxane (V) or dihydroxyisobutyl T7 silsesquioxane (Mo, W) complexes [49–51]. A chromium analog was prepared starting with chromium trioxide (see “Appendix” section for details). The reactions were carried out under inert atmosphere in dry toluene. Triethylamine was added to the reaction mixture to precipitate HCl produced during the reaction, as Et3NHCl. The vanadyl species was obtained as a monoxovanadate in the +5 oxidation state; the Cr, Mo and W species were prepared as dioxometalate species with the metal in the +6 oxidation state. The resulting materials were characterized by solution 29Si, 1H and 13C NMR, and mass spectroscopy. The OVPOSS species was also characterized by 51V NMR. The vanadium analog was also prepared via an additional method with equimolar vanadium oxytripropoxide in benzene, without the addition of triethylamine, in an effort to improve the purity and isolation of the product (see “Appendix” section for details). The results of the analyses were consistent with the expected geometries of the POSS-ligated complexes, monometallic complexes in all cases except for the vanadyl complex, which was obtained as a mixture of the mono vanadium complex and its corresponding dimer (see “Appendix” section for details).

2.1.1 Solvents

The solvents (toluene, cyclohexene, chloroform and tetrahydrofuran) were spectroscopy grade used either as received or, if necessary, after drying by passage through columns of activated alumina. We estimated the orientation polarizablities of the solvents and solvent mixtures from measurements of their indices of refraction and values for their dielectric constants calculated according to the algorithm described by Wang and Anderko [52] (see Table 3).

2.2 Spectroscopy

2.2.1 In Neat Solvents

Stock solutions of the complexes (OVPOSS, O2CrPOSS, O2MoPOSS and O2WPOSS) were prepared by dissolution of calculated weights of the complexes in the chosen solvents and dilution to yield ~8–10 μM solutions. The concentrations of the individual solutes were adjusted to an absorbance of approximately 0.5, therefore the actual concentration of the individual samples depended on the absorption coefficient of the parent POSS-ligated complex in the specific solvent. UV/Vis absorption spectra of the solutions contained in screw-capped quartz cuvettes were recorded at room temperature (23 °C) using a Shimadzu UV 3600 UV–Vis-near infrared (NIR) Spectrometer. All spectra were baseline corrected against the pure solvents and acquired with a resolution of 0.5 nm over the range 200–800 nm. Emission spectra were measured using a Horiba FluoroMax-4 Spectrofluorometer in combination with the manufacturer supplied FluorEssence software (version 3.5). The measurements were recorded with a resolution of 1 nm and averaged over five acquisitions (5 nm bandpass on both the front entrance slit and the front exit slit). The previously determined absorption maximum for each respective complex was used as the excitation wavelength; dark offset and blank subtraction corrections were applied to each spectrum. The recorded wavelength range for the spectra depended on the excitation wavelength that was used.

2.2.2 Toluene–Chloroform Solutions

The complexes were dissolved in solvent mixtures consisting of different ratios of toluene and chloroform, spanning a range from 0 to 100 vol% chloroform. Absorption and emission spectra were recorded as described above, except that the energy of the absorption maximum of each complex in toluene was used as the excitation energy for all samples of the same complex. For example, each spectrum in the series of emission spectra for OVPOSS was recorded using the excitation energy of 4.428 eV (280 nm) (see Fig. 10). As a control, we also used the energy of the absorption maxima of the chromophores dissolved in pure chloroform as the excitation energy for the emission spectra. However, we did not observe any evidence that the emission spectral profiles were influenced by the change in the excitation energy.

2.2.3 Measurements at 50 °C

O2WPOSS was dissolved in solution mixtures consisting of different volumetric ratios of toluene to chloroform as described above. The Horiba FluoroMax-4 Spectrofluorometer was equipped with a water-heated cuvette holder that permitted recording emission spectra of these samples at elevated temperature. The cuvette holder was maintained at a temperature between 48.5 and 49.8 °C during the measurements, with the aid of a programmable circulating bath, set to 50 °C. Prior to the measurement, the quartz cuvettes containing O2WPOSS dissolved in toluene–chloroform solution mixtures were preconditioned in a convection oven at ~45 °C to ensure that the actual sample would reach the desired temperature for the measurement (~50 °C) within a reasonable time frame, and to keep temperature variations between the samples at a minimum. The actual emission measurements were carried out as described above.

2.3 Modeling

In this study, the full geometry of the POSS-ligated complexes was approximated with smaller silica fragments terminated with hydrogen atoms (Fig. 3). Full geometry optimizations of the lowest singlet spin states for all systems were carried out with density functional theory (DFT) using the 1998 revised Perdue-Burke-Ernzerhof hybrid functional (PBE0) [53]. This functional, in combination with time dependent-DFT (TD-DFT), has been shown to produce fairly accurate and inexpensive predictions of optical excitations for transition metal systems [54]. A double zeta quality basis set with polarization functions was used for the main group elements (H, C, O, and Si). The Stuttgart effective core potentials with the companion double-zeta quality basis sets were used for the transition metals V, Cr, Mo and W. All equilibrium geometries were verified by Hessian calculations. Tests with bigger silica clusters gave directionally similar results (within 0.5 eV) for the oxo-metalate moiety; so we used the smaller clusters to save the computational costs for the excited state calculations. Vertical excitations to the lowest singlet excited state and “diagonal” excitations were determined using time-dependent DFT (TD-DFT [55] ) as implemented in Gaussian 09 [56]. The “diagonal” excitations (ground state minima to excited state minima) were determined by full optimization of the lowest singlet excited states.

Fig. 3
figure 3

Clusters used in the quantum chemical modeling. Left silica-ligated unioxo vanadyl, Right silica-ligated dioxo complexes of Cr, Mo, W

Full size image

We incorporated solvent effects by means of a polarizable continuum model (PCM) by placing the clusters in a cavity within the solvent reaction field [55, 57] determined by the dielectric of the solvent, i.e. chloroform (ε = 4.7113) and toluene (ε = 2.3741). Vertical and “diagonal” excitations were also obtained with the PCM approach for all systems.

3 Results and Discussion

The absorption of the POSS-ligated compounds dissolved in toluene, a nonpolar solvent with a small solvatochromic effect, exhibited absorption maxima at energies typical of those found for isolated, supported oxometalates (Table 1) [24]. Similarly, the calculated values were comparable to the experimentally measured energies, within the expected accuracy of the underlying theory and are expected to be of semi-quantitative value.

Table 1 Comparison of the measured and calculated energies for the absorption maxima of the POSS-ligated oxometalates with those of their silica-supported analogs
Full size table

The spectra of the POSS complexes (see Appendix, Fig. 10) exhibited positive Stokes shifts (Fig. 4). We included in Fig. 4 a preliminary data point for the O2WPOSS complex impregnated into MCM-41 and then dried. The point lies along the extrapolation of the trend line towards an orientation polarizability of zero, just as would be expected for an environment with the low polarizability of air. Orientation polarizabilites were calculated using Eq. 2, using the procedure described in the “Appendix” section.

Fig. 4
figure 4

Solvatochromic stokes shifts of the POSS complexes dissolved in the neat solvents. The Stokes shifts were calculated as the difference in energy between the absorption peak and the emission peak for each solution. The emission spectra measured using the absorption maxima as the excitation energy

Full size image

We used Lippert’s Relation (Eq. 1) to calculate an apparent, relative change in the dipole moment between ground and excited states Δμ = (μ G  − μ E ) for the complexes dissolved in pure solvents. Comparison of our experimentally measured values for the Group VI chromophores with those that we calculated (Table 2) suggests that the photon absorption and emission we observed correspond, as might be expected, to the transfer of an electron from the ground-state singlet state to lowest excited singlet state and its reverse.

Table 2 Measured and calculated differences in the dipole moments for the ground and Franck–Condon excited state complexes (corresponding to the right term in Lippert’s Relation), normalized, arbitrarily, to that of the O2CrPOSS complex
Full size table

The calculated Stokes shifts (see Appendix, Table 4) trend oppositely to the experimental values, likely because they involve differences of differences and are therefore very sensitive to small errors. For transition metal complexes, we expected the calculated and experimental energies for the optical transitions to be within ~0.3 eV of their calculated values [54]. The discrepancy in the absorption energy is most pronounced for W, as might be expected for a late transition metals and our use of a moderately sized basis set [54, 58]. Better agreement would require the application of more refined, and much more expensive methods that take into account electron–electron correlations.

To investigate the possibility of preferential solvation that bears on nanoscopic phase separation outlined in the introduction, we measured the solvatochromism of the POSS complexes in solvent mixtures. In this initial study, we chose to use mixtures of toluene and chloroform as the binary solvents. In each case the spectra showed shifts consistent with a toluene-rich surrounding, even at concentrations of chloroform as high as 96 mol% (Fig. 5). The data were fit well by Langmuir-like isotherms that contained one fitting parameter, an apparent equilibrium constant, K(T), to interpolate (partition) the composition (mol fraction), X, of the near-chromophore environment between toluene (Tol) and chloroform (Chl) as functions of the bulk composition of the solution (Eqs. 3–5):

Fig. 5
figure 5

Stokes shifts for all the POSS complexes fitted with Langmuir-like isotherms corresponding to the indicated values for the partition constant, K

Full size image
$$X_{Tol} = \frac{{K(T) \times X_{Tol\,bulk} }}{{X_{Chl\,bulk} + K(T) \times X_{Tol\,bulk} }}$$
(3)
$$X_{Chl} = \frac{{X_{Chl\,bulk} }}{{X_{Chl\,bulk} + K(T) \times X_{Tol\,bulk} }}$$
(4)
$$Shift = X_{Tol} \times Shift_{Tol} \times X_{Chl} \times Shift_{Chl} .$$
(5)

This analysis is a 1-parameter simplification of that employed by Bosch, et al. [59]. We were surprised that the apparent equilibrium constants were large enough to yield such strong preferentiality given that the solvents are miscible [60] (toluene and chloroform have very similar enthalpies of vaporization, 33.2 and 29.2 kJ/mol, and very similar Hildebrand solubility parameters, 18.7 and 18.3 MPa0.5). We did confirm that the solvent mixtures exhibited normal, concentration-dependent polarizabilities by demonstrating that the values of the Reichardt polarity indicator parameter [12], ET(30), for the solvatochromic benchmark, Reichardt’s dye, were linearly dependent on the orientation polarizabilties of the mixtures (Fig. 6). Therefore, we are confident that the plateau region in the solvatochromism curves for the POSS complexes in mixtures of toluene and chloroform represent preferential solvation and not some instrumental or experimental artifact.

Fig. 6
figure 6

Stokes shifts for the O2WPOSS complex dissolved in mixtures of toluene and chloroform. ET(30) for Reichardt’s dye in the same solvent mixtures shown for reference (upper line)

Full size image

We recognize that much more complicated behavior of probe molecules in mixed solvents is possible because the interactions between probe and the solvent will depend on more than merely the bulk polarizability of the solvents [14, 16, 38, 39, 59, 61, 62]. We call attention to this simple analysis because, in our examination the literature on catalysis, we have found scant reference to the use of solvatochromism for characterizing the immediate environment of the catalyst, outside of enzyme catalysis.

For the preferential solvation to play a significant role in reactions, like the upgrading of biogenic fuels, it would need to persist to the temperatures where those reactions exhibit practicable rates, for example at temperatures greater than 250 °C. Therefore, we estimated the temperature dependence of preferential solvation depicted in Fig. 5 for the O2WPOSS complex by also measuring and fitting its solvatochromism (Fig. 7) in the same solvent mixtures at 50 °C, which is about 10 °C below the boiling point of chloroform. A van’t Hoff analysis of the apparent partition constants yielded a value of about −28 kJ/mol for the apparent enthalpy of the preferential solvation. Therefore, we speculate that the preferentiality would attenuate (K = 1) at temperatures around 500 K. Because entropy of mixing would tend to cancel the preferential solvation, we considered the maximum entropy of mixing of a two-component system, which occurs for an equimolar mixture (x 1  = x 2  = 0.5). We estimate that the entropy of mixing would overwhelm the enthalpic term only at an impractically high temperature (~5000 K):

Fig. 7
figure 7

Solvatochromism of the O2WPOSS complex measured at two temperatures. Upper curve shows the isotherm corresponding to a Langmuir equilibrium constant for partitioning of unity

Full size image
$$\begin{aligned} T\Delta S &= RT\left[ {x_{1} \ln \left( {x_{1} } \right) + \left( {1 - x_{1} } \right)\ln \left( {1 - x_{1} } \right)} \right] \\ T\Delta S_{\hbox{max} } &= RT\left[ {0.5\ln \left( {0.5} \right) + \left( {1 - 0.5} \right)\ln \left( {1 - 0.5} \right)} \right] \\ &= 8.314\,{\text{J}}{\kern 1pt}\, {\text{mol}}^{ - 1}\, {\kern 1pt} {\text{K}}^{ - 1} \times 0.693 \times 5000\,{\kern 1pt} {\text{K}} \\ &= 29\,{\kern 1pt} {\text{kJ}}{\kern 1pt}\, {\text{mol}}^{ - 1} \\ \end{aligned}$$
(6)

A provocative extrapolation, given the relative contributions for enthalpy (strong) and entropy (weak), is that, the preferentiality would invert at higher temperatures, here higher than 500 K. The cybotactic region would then switch from being rich in nonpolar molecules to being rich in the polar species. Therefore the support optimization would depend on the degree of conversion (i.e., local concentration) as well as on local temperature.

In principle, therefore, for reactions like upgrading of biomass or Fischer–Tropsch synthesis, both of which necessarily involve mixtures of condensed polar and nonpolar products, the properties of a catalyst support could be tuned to optimize the near-catalyst composition of the reaction medium (Fig. 8), albeit only in condensed reaction medium because gases cannot be structured. Indeed, the recent construction of “Janus” particles [63, 64] appear to do just that.

Fig. 8
figure 8

Cartoon representation of a support tuned to encourage preferential solvation near a catalytic site (*)

Full size image

The preferential solvation that we have reported above may arise from the interaction of the solvent molecules with the organic –R groups attached to the POSS complexes (akin to the support effect mooted above) rather than an affinity towards the chromophore itself. Regardless, the results provide evidence that solvatochromism can afford an indication of changes in composition of a condensed reaction medium located near a photo-responsive moiety similar to those found in catalysts based on transition metal oxides. We expect that interesting solvent effects will arise when there is charge separation in transition states and in reaction intermediates. For example, recent quantum calculations [65] that accord with experimental reaction selectivity [66] suggest that the preference for cyclohexanone over cyclohexanol in the hydrogenation of phenol in water likely arises from the presence of polarized intermediates stabilized by a polar reaction medium.

Many spectroscopies exhibit solvatochromism. We are aware that solvatochromism has also been observed in the IR spectroscopy of surface species [67, 68]. Here, we have chosen to employ UV solvatochromism because it probes, rather directly, the localized valence electronic properties of the chromophore (Fig. 9) and provides, and, as seen from the modest accordance between theory and experiment shown in Tables 1 and 2, a challenging benchmark to test the accuracy of modern quantum chemical calculations.

Fig. 9
figure 9

Frontier orbitals of the models of oxometalates, which show that the wavefunctions involved in the UV luminescence transitions are localized on the metal centers or their immediate neighbors

Full size image

4 Conclusions

When a reaction medium consists only of highly miscible and similar components (e.g. when all are gases, or all are homologous hydrocarbons), questions about the local structure or composition of the reaction medium need not arise. In that case it is reasonable to assume that the molecules become well mixed normal to the surface of the catalyst by ordinary random motion of the fluid phase species. However, in the upgrading of biomass-derived feedstocks into fuels, the ineluctable presence of polar as well as nonpolar liquids (e.g. water, oxygenates and the product fuel molecules) raises the possibility that stratification of the reaction medium can occur because there could be preferential interaction of one of the fluid phases with the catalyst surface or with pore walls. The results described above give reason to be concerned that such cybotactic stratification can exist, even in the nearly ideal solutions we employed.

The UV solvatochromic Stokes shifts of series of POSS-tethered oxometalate compounds in different neat solvents vary linearly with a measure of the solvent polarizability but exhibit significant preferential solvation in mixed solvents. TD-DFT calculations with simple models afforded semi-quantitative predictions at no great computational cost that can be transferred to extended model systems.

Knowledge of local concentrations, or better, activities [6], of reactants and products is needed to devise accurate rate laws and to improve the design of the surfaces of contact catalysts employed in condensed reaction media. The characterization technique exemplified here provides a way to probe that information. The technique is general and can be adapted to other, in operando spectroscopies, notably polarization decay spectroscopy, which can provide information about local solution dynamics, and resonant inelastic X-ray scattering (RIXS), which could be used in optically opaque media or under conditions of high temperature and pressure, for example in the recent study of unusual solvent effects during the hydrogenation of biogenic oxygenates [7].