Pinewood with a particle size of 500 μm was chosen as the biomass feedstock because it was considered the optimal studying material with low ash content (Mendes et al., 2016). The small pinewood particle size was selected to ensure a thermally thin regime with negligible intra-particle temperature gradient during the pyrolysis process (Papadikis et al., 2009). Pinewood was dried at 105°C for 24 h in an oven before experiments and kept in a desiccator in a closed container. Nano α-Fe₂O₃ powder (MACKUN, China) and submicron α-Fe₂O₃ (FS221) (Aladdin, China) powder were calcined for 3 h at 800°C (FS054) and 3 h at 1000°C (FSm002), respectively, to obtain non-ideal spherical α-Fe₂O₃ with different particle sizes. α-Fe₂O₃ powders with dual particle size made of FS054 and FS221 (FS054-FS221) as well as FS221 and FSm002 (FS221-FSm002) were prepared by proper mixing in a mortar in 50:50 wt% ratio at room temperature. The powders FS054-FS221 and FS221-FSm002 were applied in CFP as obtained without preliminary thermal treatment.

The XRD patterns were recorded on a Bruker D8 Advance diffractometer (Bruker AXS, Germany) operating with Cu_Kα radiation in the range of 10 ≤ 2θ degrees ≤100 at room temperature under an air atmosphere. An FEI Tecnai G20 transmission electron microscope (TEM, FEI, United States) with the acceleration voltage of 200 kV and JEOL JSM-6510 scanning electron microscope (SEM, JEOL, Japan) in 5–30 keV energy range were used to characterize the morphology of catalysts. The BET measurements were carried out on a 3H-2000PS2 apparatus (BeiShiDe Instrument—S&T, China) in the static volumetric mode to estimate the specific surface area and pore sizes of the catalysts. 2–7 measurements were carried out for each sample. To evaluate quantitatively mesoporosity, the method developed by Barrett, Joyner, and Halenda (BJH method) (Barrett et al., 1951) was applied. The models of Horvath and Kawazoe (HK method) (Horvath and Kawazoe, 1983) as well as Dubinin and Radushkevich (DR method) (Dubinin, 1966) were used to estimate the microporosity.

In-situ fast pyrolysis analysis was performed with the Pyrolyser (CDS pyroprobe 5000 series, United States) in combination with Gas Chromatography-Mass Spectrometry (GC/MS) technique (7890A/5975C, Agilent technology, United States). For the non-catalytic pyrolysis experiments, 1 mg of biomass was positioned in a quartz tube sandwiched by layers of quartz wool. In-situ catalytic fast pyrolysis for vapor upgrading was studied using 1 mg of biomass and 4 mg of a catalyst properly mixed in a mortar before the final mixture was positioned into the quartz tube. As in the conventional fast pyrolysis experiments, the biomass-catalyst mixture was sandwiched between quartz wool layers. For the sake of convenience, the abbreviation “PW: catalyst” will be used for the experiments after CFP. The pyroprobe with a sample was heated at the rate of 20°C ms⁻¹–500°C. The pyrolysis process was carried out at this temperature within 20 s with a biomass-to-catalyst ratio of 1:4. Helium was used as the carrier gas with a 30 ml/min flow rate, introducing the pyrolysis vapors GC/MS system. The split ratio setting the GC/MS was 50:1, whilst the temperature of the injecting system was maintained at 300°C. The GC/MS interface temperature was held at 250°C, at a pressure of 10 psi. The chromatographic separation was performed by using 5%-phenyl-methylpolysiloxane (HP-5MS) non-polar capillary column (30 m length with 0.25 mm internal diameter). The oven temperature ramp in the gas-chromatograph was progressively increased from 50°C to 300°C at a rate of 5°C min⁻¹. The temperature of 300°C was maintained for 7 min to ensure that the column was clean from any heavy molecules. Mass spectra were obtained for m/z ratio in the range of 35–550 amu, while the GC/MS was operated at the electron ionization (EI) mode at 1459V. Reproducibility experiments were conducted three times for each sample (non-catalytic pyrolysis and catalytic pyrolysis).

Data treatment: Chromatographic peaks were identified by comparing their mass spectra to the standard spectra of organic compounds in the NIST MS library and literature data. Each product’s relative peak area percentage (%) evaluated the catalyst’s impact on the pyrolysis product distribution. However, using GC/MS as the volatile compound identification method imposes certain limitations as the relative concentrations derived from the peak area percentages are a function of the individual compound’s ionization ability. This paper assumes that the chromatographic peak area percentage of an individual compound is proportional to its concentration (first order approximation). Hence, the change in each compound’s chromatographic peak area percentage from different experiments was used to determine the catalyst impact on the relevant yield. Moreover, the detected chemical compounds were grouped according to their main functionality by using IUPAC classification. Since each experiments was carried out three times for each sample, the average peak area and its standard deviation were evaluated for each group to indicate the spread of the obtained values. The total area is in the range of 99.9–100.1% due to automatic rounding of numbers by Excel, which was the used software for data analysis.

Initially, all spherical-like α-Fe₂O₃ powders independent of the particle size were single-phase with a hematite crystal structure (Figure 1). The particle size for the powders was calculated from the TEM and SEM images (Figure 2 and Table 1). The specific surface area of powders decreases by about 8 times with the rise in the size of powders (from nano to micron size). Micro-, meso-, and macroporosities coexist in the powders. The BET measurements can quantitatively describe microporosity and mesoporosity (Table 1). The mesoporosity estimated by the BJH method is noticeably larger than the microporosity defined by the HK and DR methods. The absolute values of both mesoporosity and microporosity decrease, respectively, by about 4 times and 15–19 times with the rise in the particle size from 54 nm to 2 μm, but their ratio becomes larger for the micron-sized FSm002 catalyst (Table 1). According to the literature (Liang et al., 2022), the acidity of α-Fe₂O₃ catalysts decreases with the particle size and crystallinity increase.

In addition, the α-Fe₂O₃ catalysts with different sizes were initially characterized by XPS to describe the surface chemical states of O, Fe, C, and their surface concentrations (Tables 2, 3 and Figures 3–5). Carbon was found at the surface of all materials. The total concentration of carbon at the surface of FS054 is almost two times higher comped to FS054-FS221, FS221, and FS221-FSm002 catalysts. The three contributions can be distinguished in the C1s XPS spectra with the strongest contribution at E _BE = 284.6 eV that can be related to C-C and C-H bonds (Figure 3A, Figure 4A). This contribution is the strongest in the C1s XPS spectrum of FS054. The less intense contribution at E _BE = 286.1 eV is slightly larger for the single-sized FS054, whereas the contributions at E _BE = 288.4–288.8 eV are comparable to all catalysts. The surface concentrations of iron and oxygen in FS054 are lower than for other catalysts by 7.2–10.4 at% and 4.8–9.2 at%, respectively. The highest Fe surface concentration was revealed in the dual-sized FS221-FSm002 catalyst (29.1 at%), but it is still lower by 10.9 at% compared to the expected one according to the nominal stoichiometry. The highest oxygen surface concentration was detected for the dual-sized FS054-FS221 catalyst, and it is comparable with the nominal stoichiometry. Comparing the surface concentrations estimated by XPS with the nominal stoichiometry of α-Fe₂O_3, one could suggest that the carbon at the surface of catalysts is preferably located on the iron sites and demonstrates the following trend: the carbon concentration decreases with increasing the particle size (Table 2). However, if one looks at the re-distribution of carbon over the surface between assumed ideal oxygen and iron surface sites, FS054-FS221 catalyst demonstrates a significantly different result: carbon location over the O and Fe sites is 1:7.4, whereas it is around 1:2 for other catalysts. The [O]/[Fe] atomic ratio at the surface is always higher than the expected one for the nominal stoichiometry, and it increases gradually with the decrease in the particle size, reaching the maximum value of 2.6 for FS054. The three components can be identified in the O1s XPS spectra of initial catalysts (Figure 3B). The component assigned to the bulk oxygen at E _BE = 529.3–529.7 eV is the strongest for FS221 and the smallest for FS054. The contribution at E _BE = 530.7–531.2 eV can be associated with oxygen in the OH surface hydroxyl (Yamamoto et al., 2010), and it is larger by around 2 at% for FS054 and FS054-FS221 compared to FS221 and FS221-FSm002 catalysts (Figure 3B, Figure 4B). The contribution at E _BE = 532.1–532.9 eV related to adsorbed water molecules (Yamamoto et al., 2010) is stronger for FS054-FS221 and FS054 catalysts.

One can distinguish several contributions in the Fe2p XPS spectra of α-Fe₂O₃ with different sizes (Grosvenor et al., 2004), summarised in Table 3. The Fe2p XPS spectra of α-Fe₂O₃ with different particle sizes generally look similar for all four samples. However, all peaks in the Fe2p XPS spectrum of FS221 are slightly shifted to higher binding energy than those for FS054, FS054-FS221, and FS221-FSm002 (Figure 5), suggesting a longer retention time for physisorbed molecules during PW pyrolysis. The values for defects and for four components associated with Fe³⁺ 2p3/2 multiplets at 709–713 eV are comparable for the three FS221-based catalysts. The surface states and satellites increase gradually with the rise in the size of catalysts. The concentration of defects (at 708.0–708.3 eV) appearing near the sites with Fe ions in lower oxidation states (Heinrich and Cox, 1994), is lower by almost one order of magnitude, compared to other components, for all catalysts independently of their size and whether they are single-sized or dual-sized.

The absence of a strong peak at 25° 2θ in the XRD patterns of all catalysts after CFP, suggests the negligible formation of the crystalline carbon (Machado et al., 2013; Xu et al., 2017) or below 3 mol% considering the XRD detection limit. The peaks related to iron carbide were not detected. After CFP, only the initial α-Fe₂O₃ phase was revealed in the PW:FS054 and PW:FSm002, suggesting good chemical and phase stability of FS054 and FSm002 catalysts. The TEM image of PW:FS054 illustrates that, after the CFP, the nano-sized PW:FS054 catalyst retains its original size (Figure 6). This is very promising, in particular, for nano-sized FS054 as typically nanomaterials tend to sinter or agglomerate at intermediate or high temperatures. In contrast, FS221-containing catalysts (PW:FS054-FS221, PW:FS221, PW:FS221-FSm002) were partly converted into γ-Fe₂O₃ polymorph with both Fe²⁺ and Fe³⁺ cations compared to only Fe³⁺ cations in α-Fe₂O₃. Another important difference between these polymorphs is that the stoichiometry of γ-Fe₂O₃ can be presented as Fe(Fe_{5/3 1/3})O₄, suggesting the existence of cation vacancies. The appearance of γ-Fe₂O₃ is unexpected as α-Fe₂O₃ phase is the most stable polymorph. It is known that several factors can induce the Fe₂O₃ polymorph transformations in general: particle size, single-phase or composite system, shape of crystals, temperature, and pressure et al. (Machala et al., 2011). However, typically these transformations occur from a less stable polymorph to a more stable one. If the particle size would be the key point it is not clear why only α-Fe₂O₃ was revealed in PW:FS054 with a smaller particle size than FS221-based catalysts. One may suggest that the partial irreversible transformation of α-Fe₂O₃ to a less stable γ-Fe₂O₃ polymorph is associated with the catalytic activity of the iron cations in the reduction process and occurring of the oxidation processes during the catalytic vapor upgrading, in particular, phenols. Note that iron nanotubes were also partially converted to the less stable γ-Fe₂O₃ during the CFP and demonstrate the best performance in the phenolics reduction among other catalysts explored (Baimoldina et al., 2019). A partial reduction of iron cations followed by transformation from α-Fe₂O₃ to γ-Fe₂O₃ may be promoted due to CO and H₂ formation during the pyrolysis as well, but there is no information about the difference in CO and H₂ concentrations over different catalysts. This partial structural transformation may also be associated with ether formation in a small quantity (3.2–4.7%, Figure 7) as it was observed for the same catalysts, for which γ-Fe₂O₃ in a small quantity was detected after CFP. According to XPS, there is an enrichment in oxygen or, alternatively, a deficiency in iron cations at the surface within a few nanometers for all catalysts (Table 2). Based on these observations, the surface and the near surface region should be involved for such structural transformation to be triggered. On the other hand, these additional layers with an excess of oxygen could be built over an iron oxide catalyst through the physisorption of small, oxygen-containing fragments of biomass decomposition. When a specific critical value is reached, the phase transition from α-Fe₂O₃ to γ-Fe₂O₃ could occur.

Woody biomass comprises a mixture of biopolymers, such as cellulose, hemicellulose, and lignin at varying proportions, depending on the biomass origin and nature. The rate of decomposition and temperature range for these biopolymers vary, and, depending on the reaction atmosphere, the resulting products span a wide range of chemical groups at different proportions. Cellulose and hemicellulose are typically associated with the production of alcohols, ethers/esters, furans/furfural, aldehydes/ketones, and acids, whilst lignin is the primary source for phenolic compounds. The GC/MS chromatogram for the conventional pyrolysis of pinewood contains more than 110 peaks. The identified fast pyrolysis compounds were grouped into non-oxygenated compounds, alcohols, phenols, ethers, furans, aldehydes, furfural, ketones, acids, esters, nitrogen-containing compounds (N-containing). Non-oxygenated compounds refer to the detected oxygen-free compounds.

Initial high relative peak areas were present for the groups of phenols, ketones, acids, esters, and CO₂, whilst the overall picture is significantly different after applying the catalysts for the groups of phenols, aldehydes, and CO₂. The relative area for the phenolic compounds amounted to a total of 36.8%, originating from the decomposition of lignin. The relative peak areas of acids and ketones amounted to 9.2% and 14.2%, respectively. The formation of esters, non-oxygenated compounds, and CO₂ are associated with more complex decomposition reaction routes catalysed by the presence of inorganics existing initially in pinewood (Liu et al., 2014). In this study, it is assumed that the changes in the relative chromatographic area are proportional to the changes in concentration. However, changes in the relative peak area of the fast pyrolysis products also depend on the individual compounds’ ionisation ability; hence, this is considered a first-order approximation in the following analysis.

Figure 7 shows the effect of the mono- and dual-sized α-Fe₂O₃ catalyst particle mixtures on the product distribution of the CFP of pinewood. Phenols, aldehydes, esters, and CO₂ are the primarily affected chemical compound groups with significant variations in their relative peak area for the different catalyst sizes and combinations.

A noticeable reduction in the relative peak area of the phenolic group is observed for all applied catalysts sizes that correspond to approximately half of the initial relative peak area magnitude. No specific particle size or mixture shows a significant deviation in the phenols reduction pattern observed during CFP. All applied catalysts displayed a substantial increase in the formation of aldehyde compounds apart from the nanosize dual particle mixture of FS054-FS221, which showed no effect on promoting the formation of aldehyde compounds compared to pure pinewood pyrolysis. This is an interesting outcome, especially when correlated to the significant increase in CO₂ observed for this specific catalyst particle mixture, and it is further discussed below. The relative peak area of the esters group is significantly reduced for all applied catalysts apart from the monosized FS054, which had a minor effect in the reduction of esters compared to pure pinewood, while the appearance of ether compounds is linked to all catalysts containing the FS221 particle group. Finally, a significant increase in the formation of CO₂ is observed for all catalysts at relatively comparable levels. The FS054-FS221 mixture shows a different pattern with a relative peak area of approximately five times higher than that of the original pinewood pyrolysis. The increase in CO₂ is typically an indicator of decarboxylation and decarbonylation reactions, followed by water-gas shift reactions promoted by Fe₂O₃, associated with the conversion of various chemical groups (acids, ketones, phenols, etc.) over transition metal oxides (e.g., Fe₂O₃).

As a general observation, it is found that α-Fe₂O₃ suppresses the formation of phenolic compounds and leads to increased CO₂ formation through various reaction pathways. The impact of α-Fe₂O₃ catalysts on aldehydes, esters, and ethers shows a size or mixture-dependent behavior indicative of particle-size-dependent selectivity in the catalytic degradation reactions of fast pyrolysis vapors.

The impact of the applied catalysts with different particle sizes and mixtures within the phenolic group is shown in Figure 8. The individual phenolic compounds that have been most substantially affected are 2-methoxyphenol (i.e., guaiacol), 2-methoxy-4-methylphenol (i.e., creosol), 2-methoxy-4-vinylphenol, and 2-methoxy-4-(prop-1-en-1-yl)phenol (i.e., isoeugenol).

In terms of relative peak areas, the FS054 showed the strongest impact in the decrease of the formation of creosol, a compound that is progressively favored as the catalyst particle size increases. The FS054-FS221 and FS221 catalysts display generally similar behavior. However, the FS221 revealed a higher selectivity towards the reduction of isoeugenol, eugenol and 2-methoxy-4-vinylphenol, whereas the application of FS054 catalyst has a more substantial impact on the decrease of creosol and other phenols. Both monosized nanoparticle catalysts, FS054 and FS221, showed the highest effectiveness in decreasing the relative yield of 2-methoxy-4-vinylphenol. The FSm002 generally had the smallest impact on reducing the individual phenolic compounds, however, its overall effect is still noticeable and more intense on heavier phenolics, such as isoeugenol.

The overall impact of α-Fe₂O₃ on the catalytic decomposition of the phenolic compounds can be primarily focused on the suppression of their formation with associated release of CO₂ as shown in Figure 7. Considering that the degradation of phenolics over hematite catalysts does not lead to a significant amount of new phenolic products, it can be speculated that α-Fe₂O₃ favors the formation of radicals during lignin degradation that further enable char formation, as well as coking along the OCH₃ rearrangement pathway primarily from guaiacols (Gupta et al., 2017; Kawamoto, 2017). Although there are clear indications of the impact of catalyst particles size on the suppression of phenolic compounds, the trends tend to differ depending on molecule structure. The appearance of new phenols for nano-sized catalyst particles and mixtures, i.e., phenol for all FS221 containing catalysts and 3-methyl-phenol for FS054 and all FS221 containing catalysts, shows that nanoparticle size catalysts with increased spherical symmetry further promote demethoxylation and side-chain cracking reactions. At the same time, the formation of 3-methylphenol (i.e., m-cresol) can be attributed to the promotion of the homolytic cleavage of the O−CH₃ bond and radical-induced rearrangements (Asmadi et al., 2011; Kawamoto, 2017).

All applied catalysts showed a significant increase in the relative peak area of aldehyde compounds, apart from the nanosize catalyst mixture FS054-FS221, which showed almost no effect on the aldehyde compound group distribution. This increase is primarily associated with the formation pathway of acetaldehyde (Figure 9), a compound absent from the products of pure pinewood pyrolysis. Interestingly, acetaldehyde formation was not a promoted reaction pathway for the FS054-FS221 catalyst. Additionally, the FS054-FS221 catalyst displays a significantly higher release of CO₂ (Figure 7) than any other catalyst.

Considering the effect of α-Fe₂O₃ on the aldehyde group, it can be hypothesized that highly symmetric nano- and micro-sized catalyst particles strongly promote the formation of acetaldehyde, possibly through the cellulose decomposition pathway and, more specifically, the catalytic decomposition of levoglucosan (Zhang et al., 2012). Two possible reasons could be identified for the diverse behavior of the FS054-FS221 mixture to catalyze the levoglucosan decomposition pathway to acetaldehyde; 1) the highly asymmetrical structure of the resulting mixture with a particle diameter ratio of approximately 1:4.1, 2) an enhanced catalytic transfer hydrogenation (CTH) reaction over the carbon-rich Fe surface (Table 2) on the acetaldehyde carbonyl (Mabate et al., 2022). Hydrogen is present in the system during the biomass thermal decomposition reactions and through the water-gas shift reaction over Fe₂O₃.

For the former, this would mean that the catalyst mixture operates like a conical nanotube than a non-ideal sphere. This observation is also in line with the findings of Baimoldina et al. (2019), where highly anisotropic shape iron oxide nanotubes were incorporated in the catalytic conversion of pinewood. The high relative peak area of CO₂ for the FS054-FS221 catalyst could also indicate subsequent decarbonylation reactions of acetaldehyde to CH₄ and CO promoted due to the high effective catalyst shape anisotropy. The experimental setup in this study did not allow for the detection of light molecular weight compounds in the GC-MS, such as CH₄. However, the excessive relative peak area of CO₂ provides further evidence in this direction. The conversion of CO to CO₂ is highly favored over Fe₂O₃ catalysts through the water-gas shift reaction. Hence, a correlation between catalyst particle size ratios and final product distribution in dual-sized mixtures is evident and perhaps forms an essential parameter for controlling the final targeted chemicals.

In the latter case, the high carbon content on the Fe active site could further promote the CTH reaction on the terminal carbonyl of the acetaldehyde molecule. Acetaldehyde decarbonylation would produce CH₄ and CO as the reaction products through the CTH reaction. Subsequently, the water-gas shift reaction would proceed over the Fe₂O₃ catalyst in a similar fashion as mentioned previously.

In contrast, the FS221-FSm002 catalyst mixture with a particle diameter ratio of approximately 10:1 did not display similar behavior to FS054-FS221. It promoted the formation of acetaldehyde similar to the monosized catalysts FS054 and FS221. The anisotropic shape and the highest surface oxygen concentration of FS054-FS221 could probably play an additional role.

The general impact of α-Fe₂O₃ on the ketonic fraction is not significant, as shown in Figure 10. More specifically, all applied catalysts were found to promote the formation of 2-hydroxy-2-cyclopenten-1-one at the expense of 1,2-cyclopentanedione, possibly through a hydrogenation route. The hydrogen supplied to this reaction originates from the original biomass decomposition reactions and further enhanced through subsequent water-gas shift reactions over the iron oxide catalysts. In addition, the formation of 1-hydroxy-2-propanone, which is associated with the decomposition of the cellulosic fraction of biomass, is partially suppressed by the mono- and dual-sized nanoparticle-containing catalysts. In contrast, the mono-sized micro-particles do not display any noticeable effect.

The distribution of acidic compounds resulting from the application of various catalyst particle sizes and mixtures is shown in Figure 11. In general terms, the application of the iron oxide catalysts does not significantly affect the total relative peak area of the acidic groups. However, it has a noticeable effect on the relative distribution of the light (aliphatic acids with C ≤ 4) and heavy (aliphatic and cyclic acids with C > 4) acids. It can be generally observed that the application of the α-Fe₂O₃ catalysts results in the decrease of the heavy fraction of the acidic compound. This is significantly more intense for the monosized FS054, and FSm002 catalysts, whilst all FS221 containing catalysts display a relatively moderate behavior with only a minor decrease on the relative peak area of the heavy acids. However, no definite conclusion can be drawn since the error in the reproducibility experiments for the FS221 catalyst on the light acids is relatively large.

Generally, it can be considered that particle size does not significantly affect the distribution of light and heavy acidic compounds in the CFP of pinewood. Although the monosized catalysts, FS054, FS221, and FSm002 have significantly different specific surface areas and pore sizes (Table 1), there is no evidence to support the hypothesis that those parameters would have any major effect. The conversion of acids is probably primarily dependent on particle shape, considering the moderate effects of the dual-sized catalyst mixtures on the final distribution and their lower capacity to increase the light acid fraction. In addition, competing reactions on the more defective and anisotropically shaped spherical catalyst mixtures might further inhibit the conversion of long-chain aliphatic acids.