Raman and Fluorescence Lifetime Microscopy for Single Catalyst Particle Analysis

We prepared a series of coked samples from E-cat material and analyzed them with combined Raman and fluorescence lifetime microscopy. The catalyst material was heated up to 600 °C and dosed with 10, 24, 50, and 100 pulses of 0.1 μL hexane. Figure 1 illustrates how single particles of the E-cat sample coked with 5.0 μL hexane were analyzed with Raman and fluorescence lifetime microscopy. Figure 1a provides a schematic overview of the combined method. Pulsed laser excitation is used to probe individual coked E-cat particles. The induced Raman scattering and fluorescence signals emanating from these individual catalyst particles are split and sent to a charge-coupled device (CCD) detector for a spectral measurement, while the fluorescence decay dynamics is analyzed with single-photon avalanche diode (SPAD)-based time gating.

The initial visual inspection with the optical microscope, shown in Figure 1b, revealed large variations in catalyst particle sizes and shapes, ranging from spherical to more anisotropic. In addition, the colors of the catalyst particles ranged from light grey to black. These catalyst-particle-to-catalyst-particle variations can be attributed to the broad age distribution of the particles within the E-cat sample and compositional inhomogeneity due to a non-uniform distribution of catalyst components. Some catalyst particles are relatively fresh and active, having spherical shapes and showing high coking activity yielding graphitic black carbon on the surface (Figure 1c). On the other hand, some catalyst particles might be relatively old, damaged and deactivated, resulting in anisotropic particles showing little coking activity keeping the particles relatively light gray (Figure 1d–e).^{34, 35} The comparison of the reflection images shows that it is imperative to assess sample heterogeneity before drawing conclusions about data originating from only a few E-cat particles within a sample.

Figure 1f compares three normalized Raman spectra recorded on the E-cat sample coked with 5.0 μL hexane. The presented spectra show G- and D-bands due to ring breathing modes of graphitic sheets and their edges and defects, respectively. In addition, broad background fluorescence due to laser-induced electronic transitions in aromatic coke species is observed. Typically, the degree of graphitization of the carbon deposits is interpreted from the D-to-G-band intensity ratio.^10-14 A higher ratio implies a large number of graphitic edges and defects, suggesting that graphitic patches are relatively small compared to samples with a lower D-to-G ratio. Quantifying the areas of the bands of interest is complicated because of the broad background fluorescence, especially the broad D-band is indistinguishable from the background in spectra with relative weak Raman contribution with respect to fluorescence. Therefore, we use the Raman-to-total ratio (RTR) as a metric for the degree of graphitization of the coke instead of the D-to-G ratio. The RTR is defined as the ratio of the amplitude of the G-band above the background to the total signal at the wavelength of the G-band (Figure S2). A lower RTR can be due to a small amount of Raman scattering graphitic carbon deposits or intense background fluorescence or both, whereas a higher RTR can be due to a large amount of Raman scattering species or weak background fluorescence or both. Catalyst particle 69 has a RTR of 41 %, whereas the other catalyst particles show much weaker Raman signal (RTR=3.5 % for particle 5 and 4.5 % for particle 7). Overall, the particles coked with 5.0 μL of hexane have a RTR of 7.1±8.4 % (mean±standard deviation over 75 particles). The histogram of the RTR distribution, presented in Figure 1h, further underlines the heterogeneity of the sample.

The fluorescence decay dynamics of the signal recorded between 538 and 596 nm is illustrated with time traces in Figure 1g. The pulsed laser induces fast processes–instantaneous Raman scattering and sub-ns fluorescence indistinguishable from each other–and slower fluorescence characterized by exponential decay. Therefore, the time traces were fit to the sum of the instrument response function (IRF) describing the fast processes and two components describing fluorescence. The fluorescence components are described with the convolution of the IRF with exponential decay (Figure S3). A coked sample typically contains a wide variety of polyaromatic hydrocarbons that fluoresce with a broad range of different lifetimes.^{8, 23, 24} Here, two exponential components were sufficient to match the data accurately, so we report the weighted average fluorescence lifetime ($$ ) of the two components. Comparing $$ from catalyst-particle to catalyst-particle reveals large variations. Catalyst particle 5 exhibits a $$ of 1.16 ns, whereas catalyst particle 7 has a $$ of 2.68 ns. The fluorescence decay traces of catalyst particles 7 and 69 appear different due to the relatively large contribution of the IRF for catalyst particle 69. The average lifetime values observed for the 5.0 μL sample are distributed over 1.37±0.56 ns, as shown by the histogram in Figure 1i.

Figure 1j reveals that the RTR and the $$ are not correlated. Indeed, catalyst particles 5 and 7 exhibit a different $$ value, but a similar RTR value. Furthermore, catalyst particles 5 and 69 have a similar $$ value, but a different RTR value. This indicates that Raman and fluorescence lifetime microscopy are complimentary analytical techniques to assess the properties of coke deposits. The considerable intra-sample variations are probably due to the broad age distribution or compositional inhomogeneity of the E-cat material used for the coking experiments.

Sample-averaged Raman and Fluorescence Lifetime Microscopy

Figure 2 shows that trends in the Raman and fluorescence lifetime microscopy data for the series of E-cat materials pulse-coked with 1.0, 2.4, 5.0, and 10 μL hexane can be established when averaging over a large number of catalyst particles. The overview reflection images presented in Figure S4 show heterogeneous samples that darken with increased hexane exposure, indicating the increased coke deposition. In addition, the thermogravimetric analysis (TGA), as presented in Figure S5, showed the buildup of carbon deposits with increased hexane doses. The extended and heterogeneous coke formation was further studied with Raman and fluorescence lifetime microscopy.

The RTR histograms, presented in Figure 2a, illustrate the broad distribution of graphitization within each sample. Nevertheless, the distributions clearly shift to higher RTR with increasing hexane exposure. The average normalized Raman spectra in Figure 2b provide an alternative visualization of this trend. The D- and G-bands become more pronounced with respect to background fluorescence upon increased hexane exposure, indicating that the Raman scattering of graphitic coke becomes more dominant over the background fluorescence due to the relative lower amount of smaller aromatic-type molecules.

The average lifetime histograms, shown in Figure 2c, also show broad distributions within each sample, but a trend is discernable. With increasing degree of coking, $$ tends to increase initially and then decrease. However, the large standard deviation on $$ makes it difficult to determine if the differences in the mean $$ between the 1.0, 2.4, and 5.0 μL samples are significant. For the 10 μL sample, $$ is however clearly shorter. The averaged normalized fluorescence time traces, as shown in Figure 2d, provide another visualization of the development of the average $$ from sample to sample. The slope of the fluorescence decay of the 1.0, 2.4, and 5.0 μL samples varies slightly while the 10 μL sample clearly shows fluorescence with a shorter $$ .

The observed heterogeneity of the E-cat samples complicates the establishment of trends in the data that can facilitate the rationalization of the formation of coke. Comparing the RTR and $$ values of the coked E-cat samples in Figure S6 reveals that the two descriptors are not correlated. Averaging the Raman measurements over 75 catalyst particles brings out the increased graphitization of coke with more hexane exposure while clear developments in the fluorescence dynamics are difficult to prove. Interestingly, the TGA analysis showed only a small difference in the amount of carbon deposits for the studied samples (Figure S5). The E-cat materials coked with 1.0, 2.4, 5.0, and 10 μL hexane were composed of 0.57, 0.67, 0.70, and 1.34 % carbon deposits respectively. On the other hand, the microscopy data clearly revealed differences in the degree of graphitization of carbon deposited on the E-cat particles.

In situ Raman and Fluorescence Lifetime Microscopy

In situ spectroscopy of individual E-cat particles provides insight on the different coking stages that is not obscured by catalyst-particle-to-catalyst-particle variations. Figure 3 illustrates the tracking of single E-cat particles during hexane cracking experiments using simultaneous in situ Raman and fluorescence (lifetime) microscopy.

The heatmap of the normalized Raman spectra recorded during the coking of catalyst particle 1, presented in Figure 3a, shows that the D- and G-bands become more prominent with respect to the fluorescence background over time. Indeed, the RTR increases to 17 % during the 35 min of hexane cracking. The carbon deposits formed within the E-cat particle grow from aromatic-type molecules into graphitic-type sheets, changing from (more dominantly) fluorescent species to (more dominantly) Raman scattering species to consequently increase the RTR values. Tracking the RTR of more E-cat particles over time, as shown in Figure 3b, clearly illustrates increasing graphitization of the coke deposits for all catalyst particles. In all four runs, the RTR increases, reaching 5 % in run 2 and 4, while the RTR reaches 35 % in run 3. The graphitization rates of run 2 and 3 differ by more than a factor 13 (determined from the slope of the RTR data between 5 and 10 min). The different rates can be attributed to the differences in catalytic activity from particle to particle in the E-cat mixture.

The change in fluorescence dynamics over time during the coking of an E-cat particle is shown in Figure 3c. The fluorescence decay initially slows down over the first 1.4 minutes before accelerating, indicating a lengthening followed by a shortening of the $$ value. This kinetic behavior is observed in all four catalyst particles investigated, as depicted in Figure 3d.

The rationalization of the fluorescence lifetime trend, as illustrated in Figure 4, can yield further insights in the physicochemical events occurring within the E-cat particles. Figure 4a reveals that the fluorescence emission redshifts over time, indicating the growth of polyaromatic species. Assigning features in the fluorescence spectrum of this extremely heterogeneous sample to specific polyaromatic hydrocarbons is difficult. Not only the size but also the particular molecular structure of a polyaromatic molecule influences its fluorescence (Figure S7).¹⁷ In this work, $$ is recorded in a limited wavelength window, between 538 and 596 nm. Therefore, the $$ values represent those of polyaromatic hydrocarbons with a limited size range, most probably of 20 to 50 carbon atoms depending on the type and size of the hydrocarbon molecule. Plausible emitters are tetracene, terrylene, or a benzoid containing 10 electron sextets.¹⁷ Such limited-size hydrocarbons emit green/yellow fluorescence, while larger ones emit outside the detection window of our fluorescence lifetime setup. A simplified illustration of the formation and growth of green/yellow-emitting polyaromatic hydrocarbons in the proximity of other larger polyaromatic hydrocarbons is provided in Figure 4b. The initial limited growth from green to yellow emitters might correlate with the increased $$ value because radiative decay slows down as fluorescence redshifts.¹⁸ Further growth of some polyaromatic hydrocarbons within the measured ensemble of species advances the fluorescence redshift. These red emitters are undetectable for the lifetime measurements. In the meantime, some green and yellow emitters might maintain their molecular sizes. In time, more and more graphitic-type structures form indicated by the rising RTR (Figure 3b). The red emitters and the graphitic-type structures influence the fluorescence lifetime of remaining green/yellow emitters by Förster resonance energy transfer (FRET). FRET from an excited green/yellow emitter to nearby larger and graphitic-type structures provides an additional decay pathway for the excited state, and thereby decreases the fluorescence lifetime.^18-21 The decreasing $$ after 2 min of coking (Figure 3d) thus evidences an increasing amount of larger and more graphitic-type structures.

Combined in situ Raman and fluorescence (lifetime) microscopy provides a more detailed picture of the coking behavior of an industrially relevant E-cat material. The growth of polyaromatic species into graphitic-type carbon is illustrated by the fluorescence redshift and the increasing RTR. The fluorescence lifetime measurements indicate the prolonged presence of smaller fluorescent hydrocarbons of which the emission is most probably quenched progressively by nearby larger polyaromatics that act as FRET acceptors. The technique developed here was also used to investigate the regeneration of coked E-cat particles to reveal the breaking of larger graphitic coke structures and the cleansing of the catalytic surface, as presented in Figure S8. During the regeneration, the D- and G-bands become less prominent with respect to the fluorescence background, indicating the breaking down of larger, graphitic structures into smaller polyaromatic hydrocarbons. These species shrink over time, indicated by a fluorescence blueshift. Furthermore, $$ lengthens during the regeneration. Here, the clearance of the catalyst surface, due to the combustion of coke molecules, might also influence the lifetime of green/yellow emitters due to FRET. Additional decay pathways for excited green/yellow emitters via FRET to nearby larger structures become less and less available, increasing $$ .