Ru/CeO₂ catalysts for supercritical water gasification of glycerol: stability, activity, and coke hydrothermal oxidation

Abstract

Cerium oxide (CeO₂) was investigated as a support for Ru nanoparticles in the continuous supercritical water gasification (SCWG) of glycerol as a model biomass compound. The stability of pure CeO₂ under reductive SCW conditions was first assessed in both batch and continuous modes. In batch, CeO₂ readily underwent phase transformation to CeCO₃OH, implying reduction of Ce⁴⁺ to Ce³⁺, leading to severe loss of surface area. The formation of CeCO₃OH was completely prevented in continuous flow due to the flushing of CO₂ and reductive compounds before cooling down to a temperature thermodynamically favouring CeCO₃OH. Although pristine CeO₂ showed low activity in SCWG, thermal treatment strongly influenced product distribution, with CO disappearing in favour of C₃H₈ after a reductive pretreatment. When Ru nanoparticles were supported on thermally stabilized CeO₂ (calcined at 700–900 °C, 0.11–1.00 wt% Ru), the catalysts displayed high efficiency in SCWG of glycerol, yielding CO₂, CH₄, and H₂ as the main products. These Ru/CeO₂ catalysts achieved apparent conversion rates up to three times higher than a commercial 5 wt% Ru/AC benchmark, while maintaining similar selectivity trends. Both Ru/CeO₂ and carbon-based catalysts exhibited similar decreasing activity trends under operation. Post-reaction analysis revealed some carbon deposition on Ru/CeO₂, likely the main cause of the observed loss of activity, which could be effectively removed either by thermal oxidation in air at 300 °C or by in situ hydrothermal oxidation with H₂O₂ at 110 °C. Overall, Ru/CeO₂ emerges as a durable and regenerable alternative to Ru/carbon for SCWG applications.

---

Introduction

Climate change, driven by anthropogenic greenhouse gas (GHG) emissions, is intensifying extreme weather events, threatening ecosystems, agriculture, and economies.^1,2 The Paris agreement seeks to limit global warming below 2 °C, ideally 1.5 °C, requiring rapid GHG reduction.³ Some sectors, such as transportation and heavy industry, are hard to decarbonise and rely heavily on fossil fuels,⁴ necessitating a shift to renewable chemical energy carriers. Currently contributing to 9.8% of the global energy supply,⁵ biomass offers a promising alternative, enabling carbon-neutral energy while valorising organic waste streams. However, traditional biomass utilization faces hurdles, including low energy density, land-use conflicts, and high water content.^6,7

Hydrothermal gasification (HTG), particularly supercritical water gasification (SCWG), efficiently converts wet biomass into methane or hydrogen-rich gas under high-temperature, high-pressure conditions.^8–14 Unlike conventional methods, SCWG eliminates the need for drying the feedstock, making it ideal for processing sewage sludge, wet agricultural residues, municipal waste, and algae. By enhancing waste valorisation and clean energy production, SCWG can help reduce society's dependence on fossil fuels.¹⁴ SCWG converts wet biomass into gas through various competing reactions, with composition dependent on process parameters and catalyst used. Operating at 380–700 °C and pressures above 221 bars, SCWG gasifies diverse organic feedstocks^8–12 and separates inorganic components due to their low solubility in supercritical water (SCW).^15–17 Catalytic SCWG (cSCWG), typically operated at 374–450 °C, yields a methane-rich gas through the use of Ru- or Ni-based catalysts while allowing high conversion rates.^9,10,18

State-of-the-art catalysts for cSCWG are ruthenium nanoparticles on carbon-based supports, typically activated carbon.^10,19,20 Nickel, despite its lower price, exhibited poor stability under these conditions, suffering from sintering or leaching.^21–24 Ru nanoparticles supported on carbon nanofiber (CNF) showed outstanding activity, and above 1.5 nm diameter, the particle size showed negligible impact on the intrinsic activity of Ru, i.e. no effect on turnover frequency (TOF).²⁵ Ru particles below 1.5 nm diameter proved intrinsically more active but also suffered from severe deactivation, as opposed to larger particles. Interestingly, a clear correlation between TOF and Ru surface loading was found.²⁵ These catalysts show, however, fast deactivation in the first hours of testing due to some particle size growth (“sintering”) and coking, with an exponential decay stabilizing towards a steady-state TOF. The size of supported Ru particles have been shown to always stabilise to the same range of 2–3 nm for CNF, AC, or ZrO₂, even at high loadings, after continuous cSCWG tests using model feedstock.^20,25 Aside of catalyst poisoning, coking remains the main cause of deactivation. Regeneration of Ru/AC catalysts is possible via mild hydrothermal oxidation with H₂O₂ to remove sulphur and/or coke, but it is still challenging because selective coke removal inevitably causes partial damage to the carbon support.^26–28 Ru remains the best active phase for cSCWG in all aspects, but its high cost represents a non-negligible expenditure. The yearly OPEX linked to the catalyst regeneration, its partial replacement and that of the sulfur scavenger represents ca. 8% of the CAPEX of a 3 t h⁻¹ plant for wet biomass or wastes such as sewage sludge or digestates.²⁹ In order to reduce this cost, beyond its high activity, a catalyst with a high stability and regenerability is of utmost importance.

Catalysts using oxides as support have so far systematically shown poor TOF and/or poor stability in continuous-flow SCWG, when compared to carbon-supported catalysts.^10,20,30,31 Stable oxides typically have low specific surface area under SCWG conditions (reducing conditions), and they are significantly surpassed by carbon with its high surface area, good stability and low cost.³¹ Besides these advantages, Ru nanoparticles show much higher performance when supported on carbon, compared to any oxide investigated so far.

CeO₂ is known for its oxygen storage capacity (OSC) allowing oxygen donation to supported metal nanoparticles, which limits coke formation.³² The operating temperature of OSC is highly dependent on the heat treatment conditions, morphology, and porosity of CeO₂, but is typically in the range of 300–500 °C.³³ Several recent studies have shown a clear link between the OSC of the support and the catalyst activity when CeO₂-based supported nanoparticles were used, for instance in the water gas shift reaction (WGS)^34–38 or reverse WGS.³⁹ The presence of noble metal nanoparticles such as platinum was proved to enhance the reducibility of a CeO₂-based support.⁴⁰ The OSC of CeO₂ can be tuned (increased or decreased) by adding to it various elements (“doping”), such as Zr, La, or Fe. The choice of the dopant was found to severely affect the selectivity of a catalyst,⁴¹ and to influence adsorption and activation of gases on the support.⁴²

The low solubility of cerium oxide in pure SCW was reported from thermodynamic calculations, and its potential stability was experimentally confirmed,^21,22 but the nature of the stable phase strongly depends on pH, oxygen fugacity and temperature.²¹ While physical stability of CeO₂ nanoparticles under reductive supercritical water was proven,⁴³ the partial reduction of Ce⁴⁺ to the Ce³⁺ phase (formation of CeCO₃OH), has been reported⁴⁴ and is expected to be the cause of a major loss of specific surface area and porosity.

This paper aims at exploring the potential of CeO₂ as a support for Ru nanoparticles for the continuous cSCWG. First, the stability of CeO₂ under reductive cSCWG conditions was investigated, both in batch and continuous mode. Then, the most stable oxides were used to prepare Ru/CeO₂ catalysts with different loadings, in a similar range as those determined as optimal for CNF as a support.²⁵ Finally, their performances were then evaluated in the cSCWG of glycerol and compared with the benchmark catalysts Ru/CNF and Ru/AC. Coke deposits formed on the spent catalyst were studied and the regeneration capability of such class of catalysts was assessed as well.

Results and discussion

Ceria stability in SCWG: batch vs. continuous

The X-ray powder diffraction (XRD) patterns recorded for the pristine CeO₂ sample and the ones after the performed SCWG of glycerol (batch and continuous flow experiments) at 430 °C and 270 bar are shown in Fig. 1. The raw CeO₂ possesses a set of diffraction lines at 2θ = 28.5°, 33.1°, 47.5°, 56.3°, 59.1°, 69.4°, 76.6° and 79.1°, assigned to the (111), (200), (220), (311), (222), (400), (311), (420) and (113) planes typical of pure CeO₂ in the face-centred cubic fluorite structure with Fm-3m space group (JCPDS 34-0394).⁴⁵ After being exposed to pure supercritical water for 30 min (in the batch reactor), the material did not suffer from any loss of physical integrity nor from any significant phase evolution. However, an increase of crystal size from 18 to 30 nm was observed resulting in a decrease in surface area from 53 to 17 m² g⁻¹ (Fig. S1 and Table 1).

Fig. 1 XRD patterns of CeO₂ before and after SCWG batch and continuous experiments.

Table 1 Textural properties of CeO₂ after various treatments

Sample	SCW/SCWG conditions	SBET [m^2 g^−1]	Vtotal [cm^3 g^−1]	Crystallite size^a [nm]
a Crystallite size of CeO2 evaluated via Rietveld refinement.b Note that the mechanical strength (not measured) of the pellet after the test without buffering has significantly increased, as evidenced during crushing of the pellets prior to XRD analysis.
CeO2	N/A	53	0.098	18
CeO2_SCW_batch	H2O, 30 min, 430 °C, 270 bar	27	0.095	30
CeO2_SCWG_batch	10 wt% glycerol, 30 min, 430 °C, 270 bar^b	7.0	0.028	30
CeO2_SCWG_batch_pH = 7	10 wt% glycerol in phosphate buffer, 30 min, 430 °C, 270 bar	16	0.050	28
CeO2_SCWG_continuous	10 wt% glycerol, 15 g min^−1, 3 h, 430 °C, 270 bar	20	0.095	31

---

The oxygen fugacity of the reaction media greatly impacts the stability of the cerium phases under the conditions applied,²¹ therefore tests with 10 wt% glycerol in water were also performed. During these experiments under reductive conditions, the CeO₂ sample evolved rapidly, and already after 30 min, a new highly crystalline phase appeared, aside the cubic fluorite CeO₂ structure. The most intense set of reflections of this new phase were at 2θ = 17.7°, 24.5°, 30.4°, 35.9°, 43.3°, 44.0°, 47.1°, 53.8°, 57.4° and 61.2°, which can be assigned to the (002), (030), (032), (004), (330), (034), (332), (062), (334) and (036) planes, respectively, typical for the CeCO₃OH phase crystalised in the hexagonal space group P-6 (ICSD 238537).⁴⁶ The formation of this CeCO₃OH phase involves the reduction of Ce⁴⁺ to Ce³⁺. Detailed phase analysis using Rietveld refinement of the XRD data was performed with GSAS-II software (Fig. S2 and Table S1) and good agreement was obtained for the two-phase model (CeO₂ and CeCO₃OH). Based on performed calculations, this sample consists of 44 wt% CeO₂ and 56 wt% CeCO₃OH. In parallel to this partial change of phase, a significant reduction in porosity of the spent material was observed (Table 1). The BET surface area (S_BET) and total pore volume (V_total) obtained from N₂ physisorption analysis dropped drastically from 53 to 7.0 m² g⁻¹ and 0.098 to 0.028 cm³ g⁻¹, respectively. The phase change was also associated with a significant increase of the mechanical strength (not measured) of the pellet as evidenced during crushing of the pellets prior to powder XRD analysis. The observed phase transformation is consistent with the results from Guan et al., who also noted a phase transformation of ceria (support of Ru phase) after SCWG of phenol in a batch reactor.⁴⁴ The formation of water-soluble Ce³⁺ species (CeO⁺, CeOH²⁺) have been predicted by Jocz et al. at concentrations higher than ≈1 mmol kg⁻¹ to occur under reductive supercritical water conditions below pH 6 at 380 °C and 300 bar. This pH limit (recalculated under process conditions) declines with increasing temperature and decreasing pressure, e.g. pH 4 at 450 °C and 300 bar or 380 °C and 220 bar.²¹ During the gasification of glycerol under batch conditions, the pH of the solution decreased from 7 to 4. To evaluate the effect of pH on the formation of this phase, an attempt was made to stabilise CeO₂ by controlling the pH of the glycerol feed solution. For this purpose, a phosphate buffer was added. Note that despite of this buffer, the pH decreased from 7 to 6 (measured before and after the test at ambient conditions) after the SCWG experiment under analogous conditions. The post-test XRD analysis (Fig. S3) shows that the increase of the pH limited the appearance of the undesired phase but did not prevent it (sample composition: 70 wt% of CeO₂ and 30 wt% of CeCO₃OH). The control of the pH also led to a smaller decrease in porosity compared to the experiment conducted without pH stabilization (S_BET = 16 vs. 7.0 m² g⁻¹ and V_total = 0.050 vs. 0.028 cm³ g⁻¹). This supports the hypothesis that a dissolution-recrystallization process is favoured at lower pH, inducing coarsening (“sintering”) of CeO₂ and resulting in higher concentration of Ce³⁺.²¹

Thermodynamic calculations of CeO₂ stability under hydrothermal conditions reveal that at pH = 7, a pressure of 300 bars and under reductive conditions, Ce³⁺ species (CeO⁺, CeOH²⁺, Ce³⁺) start to dominate over Ce⁴⁺ species upon cooling from 370 °C to 150 °C.²¹ This is in phase with the rather low thermal stability of CeCO₃OH, which depends strongly on the conditions it is exposed to.⁴⁷ Under reductive (H₂) static gaseous conditions, decomposition starts at 248 °C until 455 °C to form CeO₂, with 75% of mass loss occurring before reaching 400 °C.⁴⁷ Note that under inert conditions, the range of the decomposition temperature of CeCO₃OH increased to 336–467 °C. Based on these two facts, the hypothesis can be drawn that the formation of CeCO₃OH observed after batch tests occurs upon cooling down at the end of the batch test. Indeed, during the cooling, CeO₂ is exposed to conditions that will favour the formation of the hydroxy carbonate, that is, CO₂ in excess, a temperature lower than 400 °C, and reducing conditions, thanks to the presence of unreacted glycerol and to its decomposition products such as H₂ and CO, but also acrolein, propionaldehyde, acetaldehyde and allyl alcohol.⁴⁸ The very different conditions CeO₂ is exposed to during the batch vs. continuous tests are displayed in Fig. 2. In this Figure, the lines represent the boundaries delimiting the conditions below which one or several aqueous Ce³⁺ species are likely to be equilibrated at a concentration higher than ≈1 μmol kg⁻¹_H2O, for different temperatures. Above these lines, CeO₂ dominates and these water-soluble Ce³⁺ species have concentrations below the ppm range. The black crosses, representing the conditions met during a batch test, clearly show that during all steps Ce reduction and the formation of water-soluble species were favoured. During a continuous test (grey points), the material would not be exposed to such unfavourable conditions, i.e. CeO₂ is expected to be stable. However, when glycerol is fed continuously, this limit is likely reached. Note that the effect of batch vs. continuous flow on the pH is complicated to assess, but a lower pH is expected during the batch test.

Fig. 2 Calculated f_O2 ‐ pH diagrams for the Ce–H₂O system. f_O2 is expressed in bar, and pH is in log₁₀(mol kg⁻¹_H2O). Boundaries indicated for different temperature delimits the conditions below which one or several aqueous Ce³⁺ species are likely to be equilibrated at a concentration higher than ≈1 μmol kg⁻¹_H2O. Conditions applied during batch experiments are presented in black: a/step during heating up, b/steady reaction, c/step during cooling. Conditions applied during continuous test are in grey: a₁/step during heating up, a₂/steady temperature, under pure water, b/steady reaction with glycerol, c₁/flushing step with water, c₂/step during cooling. Adapted from.²¹

To test this hypothesis, experiments with pure CeO₂ oxide under continuous flow of a 10 wt% glycerol solution were performed at 430 °C, 270 bar and 15 g min⁻¹ for 3 h. After the experiment was completed, the catalyst bed was rinsed with water for 45 min (at SCW conditions), after which the system was cooled under water (typical steps followed for any continuous test) (see grey crosses in Fig. 2). In this case, a significantly smaller decrease in both BET surface area and pore volume was observed after the continuous SCWG of glycerol (20 m² g⁻¹ and 0.095 cm³ g⁻¹), similar to the decrease observed after the batch test with pure SCW (27 m² g⁻¹and 0.095 cm³ g⁻¹), and the crystallite size were found to be nearly identical, at 28 and 30 nm, respectively. Most importantly, the analysis of the material after the test revealed the total absence of CeCO₃OH (Fig. 1). These observations are in line with the hypothesis that CeCO₃OH formed predominantly upon quenching the batch reactor, as depicted in Fig. 2. Therefore, the primary reduction in the S_BET of pristine CeO₂ (from 53 down to 20–27 m² g⁻¹) is probably due to the harsh environmental conditions of SCW. Further reduction of the surface area (from 20–27 down to 7.0 m² g⁻¹) is likely linked to the formation of large CeCO₃OH crystals during the gasification of glycerol (only during batch experiments, when a solid sample remains in the reaction medium during cooling). These results, however, do not exclude the possibility that some CeCO₃OH was formed during continuous flow reaction and then decomposed in the presence of pure water. The concentration of cerium ions found in the process water produced using 0.44%Ru/CeO₂_700 at 401 °C, 270 bar and a flow rate of 7.2 g min⁻¹ of 10 wt% glycerol was 4.6 × 10⁻⁸ mol_Ce kg⁻¹, matching the values predicted by thermodynamic calculations.²¹

With the aim to increase the stability of the CeO₂ support (before SCWG experiment), the effect of stabilization of CeO₂ by thermal treatment at 500, 700, and 900 °C was investigated. The XRD patterns of the calcined powders show the expected behavior, with narrower and more intense peaks being observed as the calcination temperature increases (Fig. S4). Thus, the coherence length estimated via Rietveld refinement increased from 18 nm (for the unmodified sample) to 20, 31, and 82 nm after treatment at 500, 700, and 900 °C, respectively. Accordingly, specific surface area and total pore volume decrease due to particle coarsening (Table 2 and Fig. S5). For CeO₂ treated at 500 °C, no changes were observed with respect to pristine CeO₂ (S_BET = 53 m² g⁻¹ and V_total = 0.098 cm³ g⁻¹), but after calcination at higher temperatures (700 and 900 °C), the S_BET values dropped to 30 and 3.0 m² g⁻¹, and V_total to 0.068 and 0.022 cm³ g⁻¹, respectively.

Table 2 Evolution of surface area and total pore volume of pristine CeO₂ and after various thermal treatments and SCWG experiments

Sample	SBET [m^2 g^−1]	Vtotal [cm^3 g^−1]	Δ^spent,fresh^a (SBET)[%]	Δ^spent,fresh^a (Vtot)[%]
			Batch^b/continuous^c SCWG
a Fresh and spent refer to the sample before and after the SCWG test, respectively.b Batch reaction conditions.c Continuous conditions.d As received CeO2.
CeO2^d	53	0.098	−87/−62	−71/−3
CeO2_500	30	0.098	n.d.	n.d.
CeO2_700	30	0.068	n.d./−40	n.d./−1
CeO2_900	3.0	0.022	0/−3	0/0

---

As expected, the performed CeO₂ calcination proved to be beneficial to the stability of the catalyst in SCWG process. As shown in Table 2 and S3, increased calcination temperature decreased the loss in the specific surface area and pore volume after SCWG experiments (more noticeable for S_BET then V_total). For example, for CeO₂ annealed at the highest investigated temperature (900 °C) only minimal changes were observed (S_BET = 3.0 m² g⁻¹ vs. 2.9 m² g⁻¹, while V_total was identical in both cases, V_total = 0.022 cm³ g⁻¹). The higher structural stability of the material can be explained by the larger particles it is composed of, resulting in a much lower surface energy contribution to minimize via coarsening. It should be noted, however, that this stability is accompanied by a very low specific surface area, which is partially explained by the high density of CeO₂ (7.22 g cm⁻³). Additionally, the high-temperature calcination process is associated with a much less pronounced formation of CeCO₃OH phase under batch conditions (Fig. S6), which probably is mainly caused by the lower surface area of CeO₂ exposed to reactive conditions. Based on Rietveld's refinements (Fig. S2C), it was estimated that this sample consists of as much as 86 wt% CeO₂ and only 14 wt% CeCO₃OH (which is more than a threefold decrease in the mass of CeCO₃OH compared to the non-calcined CeO₂ sample).

Ru/CeO₂ performance

The efficiency of Ru/CeO₂ catalysts (Ru loading 0.11, 0.25 and 1.00 wt%) in continuous SCWG of 3 wt% glycerol was studied at a pressure of 275 bar, 400–430 °C and weight hourly space velocities (WHSV) in the range of 1100–3400 g_glycerol g_Ru⁻¹ h⁻¹. Film and internal mass transfer limitations were assessed using Mears' criterion and the Weisz–Prater criterion and were proven to be negligible (see details in SI). Since the considered materials differed significantly in Ru loading, their efficiency was expressed as the apparent carbon conversion rate (r_Xc), based on the TOC analysis of the process water, per gram of active phase (Ru) and per unit of time (h), Fig. S7 and Table 3. The 0.11%Ru/CeO₂_900 and 0.25%Ru/CeO₂_700 catalysts showed calculated r_Xc, averaged over the steady-state reached at specific conditions, ranging from ca. 470 ± 30 to 1290 ± 60 g_C g_Ru⁻¹ h⁻¹ and from 430 ± 30 to 990 ± 50 g_C g_Ru⁻¹ h⁻¹, respectively. Importantly, in both time intervals, in which the same conditions of the SCWG experiment occurred (400 °C, 10 g min⁻¹), the 0.11%Ru/CeO₂_900 catalyst showed a very high carbon conversion of 96 ± 3% and 98 ± 5% (Time On Stream (TOS) = 2.8 and 6.4 h, respectively), which resulted in almost identical apparent carbon conversion rates of 1270 ± 40 and 1290 ± 60 g_C g_Ru⁻¹ h⁻¹. In turn, 0.25%Ru/CeO₂_700 tested under the same conditions exhibited a slightly lower conversion value, i.e. 78 ± 4% (TOS = 3.8 h) and 71 ± 3% (TOS = 6.2 h), which corresponded to 990 ± 50 and 910 ± 40 g_C g_Ru⁻¹ h⁻¹, respectively. Therefore, it can be stated that both tested catalysts exhibit high stability indicated by good performance repeatability for changing SCWG conditions over time. Moreover, it should also be noted that for these catalysts, regardless of the reaction conditions, the C-based glycerol conversion always reached values above 70% (Table 3), therefore the calculated apparent conversion rate values were not determined in the kinetic regime. In contrast, the sample with the highest Ru loading (1.00 wt%) showed a much lower apparent conversion rate (ca. 220 ± 10 to 460 ± 10 g_C g_Ru⁻¹ h⁻¹, depending on SCWG conditions) compared to the other tested Ru/CeO₂ catalysts with lower Ru content. However, despite the lower efficiency, it did not show any signs of deactivation in the tested timeframe, as for the catalysts with lower loading.

Table 3 Conditions of continuous SCWG of 3 wt% glycerol and performance of Ru/CeO₂ catalysts

Catalyst	TOS [h]	Temperature [°C]	Flow rate [g min^−1]	WHSVRu [gglycerol gRu^−1 h^−1]	Conversion [%]	Apparent C conversion rate [gC gRu^−1 h^−1]
0.11%Ru/CeO2_900	1.0	401	7.2	2380	90 ± 1	840 ± 10
	2.8	401	10.3	3390	96 ± 3	1270 ± 40
	5.0	427	3.7	1220	98 ± 6	470 ± 30
	6.4	401	10.2	3370	98 ± 4	1290 ± 60
0.25%Ru/CeO2_700	1.5	401	7.1	2260	89 ± 4	790 ± 30
	3.8	400	10.2	3270	78 ± 4	990 ± 50
	5.2	432	3.7	1170	93 ± 6	430 ± 30
	6.2	401	10.3	3290	71 ± 3	910 ± 40
0.44%Ru/CeO2_700	1.5	400	7.5	2010	99 ± 1	780 ± 10
	2.9	400	12.0	3230	99 ± 1	1260 ± 10
	4.5	400	7.5	4060	83 ± 7	1310 ± 10
	6.0	400	7.5	6870	28 ± 6	740 ± 10
1.00%Ru/CeO2_700	1.0	399	7.4	2270	52 ± 1	460 ± 10
	2.0	400	3.7	1140	50 ± 1	220 ± 10
	3.3	400	10.3	3180	35 ± 1	430 ± 10
	4.5	430	4.0	1230	57 ± 2	270 ± 10
	5.5	401	10.3	3160	34 ± 1	420 ± 10

---

In turn, 0.44%Ru/CeO₂_700 was tested twice. The first measurement was performed at 400 °C, 275 bar pressure and varying feed glycerol concentrations (3, 6 and 10 wt%), Fig. S7D. At low glycerol content during the SCWG run (3 wt% glycerol, 7.5 g min⁻¹), the apparent carbon conversion rate was 780 ± 10 g_C g_Ru⁻¹ h⁻¹, and after increasing the glycerol mass flow rate to 12.0 g min⁻¹, a value of 1260 ± 10 g_C g_Ru⁻¹ h⁻¹ was achieved. A similar value (1310 ± 10 g_C g_Ru⁻¹ h⁻¹) was noted after increasing the glycerol feed concentration to 6 wt% and at a mass flow rate of 7.5 g min⁻¹. However, further increasing the glycerol solution concentration to 10 wt%, while maintaining the feed mass flow rate, resulted in a significant drop in conversion (to 28 ± 1%), which was reflected in a decrease in the apparent conversion rate to 740 ± 10 g_C g_Ru⁻¹ h⁻¹. The second measurement was performed under constant SCWG conditions (i.e., 400 °C, 275 bar, 5 g min⁻¹, 3 wt% glycerol), to determine the stability of this catalyst over time on stream (Fig. S7E). Under the applied conditions, a noticeable decrease in the apparent carbon conversion rate was observed (from 2200 ± 60 to 1340 ± 20 g_C g_Ru⁻¹ h⁻¹), indicating a similar degree of deactivation as that reported for CNF-supported Ru nanoparticles;^25,48 both points being further discussed in the next sections. Additionally, the apparent carbon conversion rate based on the gasification efficiency values (r_GEc) depicted in Fig. S8 showed similar numbers to those discussed in Fig. S7. However, the apparent deviations between the recorded points at each time interval were caused by the low gas flow and the dilution of the gas effluent by N₂ used to prevent GC detector peak saturation by H₂ as well as to increase the gas pressure for low gas production experiments.

For all tested Ru/CeO₂ catalysts, the main reaction products were CO₂, CH₄, and H₂. With the change of SCWG process conditions (and the extension of reaction time), the gas composition changed slightly. Over 0.11%Ru/CeO₂_900, the H₂ content in the produced gas increased from 17 ± 1 to 31 ± 2 vol% at the expense of CH₄, which decreased from 47 ± 3 to 38 ± 3 vol%. A similar decreasing trend of produced CH₄ was observed with 0.25%Ru/CeO₂_700 as well as 1.00%Ru/CeO₂_700 (drop from 33 ± 2 to 22 ± 2 vol% and from 28 ± 1 to 19 ± 2 vol%, respectively). During this time, the H₂ concentration in the gas mixture also changed, with a clearly visible upward trend (31 ± 2 to 42 ± 2 for 0.25%Ru/CeO₂_700 and 38 ± 2 to 44 ± 2 for 1.00%Ru/CeO₂_700). For all tested catalysts and SCWG conditions, the concentration of CO₂ produced remained at a similar level and ranged from 32 ± 2 to 38 ± 2 vol%, while the CO content was usually below 1 vol%.

Fig. 3 shows the apparent carbon conversion rate over Ru/CeO₂ samples (Ru loading: 0.11, 0.25 and 1.00 wt%) determined after 6 h of operation at a pressure of 275 bar, a temperature of 400 °C and a flow rate of 10 g min⁻¹ of 3 wt% glycerol in water. This comparison also included the results obtained under the same temperature and pressure but lower feed mass flow rate, i.e. 5 g min⁻¹ for 0.44%Ru/CeO₂_700 and 6 g min⁻¹ for the research benchmark 5.00%Ru/CNF catalyst, which showed the highest performance in our previous studies,²⁵ and the commercial benchmark 5.02%Ru/AC catalyst used at pilot scale.⁴⁹ Among the studied Ru/CeO₂ samples, the highest apparent carbon conversion rate was observed for 0.44%Ru/CeO₂_700 and 0.11%Ru/CeO₂_900, for which value of 1340 ± 20 and 1230 ± 60 g_C g_Ru⁻¹ h⁻¹ were obtained, respectively. For other CeO₂-supported catalysts, lower apparent conversion rate values were achieved, i.e. 910 ± 40 g_C g_Ru⁻¹ h⁻¹ for 0.25%Ru/CeO₂_700 and 420 ± 10 g_C g_Ru⁻¹ h⁻¹ for 1.00%Ru/CeO₂_700.

Fig. 3 Apparent C conversion rate per gram of Ru (red bars) and per catalyst bed volume (blue bars). The dotted blue line indicates the volumetric conversion rate determined for CeO₂_700 support. Conditions of SCWG: 3 wt% glycerol in water, p = 275 bar, T = 400 °C, feed flow rate of 10 g min⁻¹ for 0.11%Ru/CeO₂_900, 0.25%Ru/CeO₂_700, 1.00%Ru/CeO₂_700 samples, 5 g min⁻¹ for 0.44%Ru/CeO₂_700 and 6 g min⁻¹ for 5.00%Ru/CNF and 5.02%Ru/AC. Apparent carbon conversion values and WHSV in g_C g_Ru⁻¹ h⁻¹ are given in the figure. The asterisks indicate a possible apparent conversion rate underestimation because the conversion was not low enough.

For comparison, the commercially available 5.02%Ru/AC catalyst reached 510 ± 20 g_C g_Ru⁻¹ h⁻¹, which is slightly higher than for 1.00%Ru/CeO₂_700, and at the same time almost two and three times lower than for 0.25%Ru/CeO₂_700 and 0.11%Ru/CeO₂_900 as well as 0.44%Ru/CeO₂_700, respectively, which clearly indicates the high efficiency of the investigated catalysts. In turn, the benchmark 5.00%Ru/CNF catalyst turned out to be only slightly more effective compared to Ru/CeO₂, achieving an apparent C conversion rate of 1490 ± 30 g_C g_Ru⁻¹ h⁻¹. As the spent catalysts had different specific surface areas, the apparent carbon conversion rate was also calculated based on the measured BET values. Considering the Ru/CeO₂ systems, the highest value of 0.49 g_C m⁻² h⁻¹ was achieved for 0.11%Ru/CeO₂_900, while for the other catalysts with higher Ru loading, 0.35 g_C m⁻² h⁻¹ was reached for 0.44%Ru/CeO₂_700 and 1.00%Ru/CeO₂_700, and only 0.21 g_C m⁻² h⁻¹ for 0.25%Ru/CeO₂_700. It is worth mentioning that in the case of pure CeO₂_700 support this value was only 0.02 g_C m⁻² h⁻¹, which indicates that the influence of the CeO₂ matrix itself on the SCWG process is virtually insignificant. On the other hand, for carbon-supported catalysts these values were calculated to be 0.25 g_C m⁻² h⁻¹ (5.00%Ru/CNF) and 0.31 g_C m⁻² h⁻¹ (5.02%Ru/AC). Additionally, the tested supported catalysts possess also different bulk densities, which were determined by the tap density method. The dry catalyst was placed into a narrow glass tube, which was tapped until the catalyst bed volume did not change. The bulk density calculated in this way was 2.33 g cm⁻³ for all Ru/CeO₂, while much lower values of 0.60 g cm⁻³ and 0.20 g cm⁻³ were noted for 5.02%Ru/AC and 5.00%Ru/CNF, respectively, which is inversely proportional to the volume of the catalyst bed in the reactor. At elevated pressure and temperature, reactor volume is costly; thus, reducing it, for example by employing high-density catalysts, can significantly lower the capital expenditure of a cSCWG process. Therefore, the performance of the studied catalyst was calculated as an apparent conversion rate of glycerol per catalyst volume (Fig. 3, blue bars). Among the CeO₂ supported samples, the highest value of 9.8 g_C cm_cat⁻³ h⁻¹ is achieved with the 1.00%Ru/CeO₂_700 sample, which was previously considered to be the least promising when considering the apparent conversion rate per gram of Ru. The 0.44%Ru/CeO₂_700 and 0.11%Ru/CeO₂_900 performed slightly worse, reaching values of 7.8 and 7.2 g_C cm_cat⁻³ h⁻¹, respectively, while the weakest system turned out to be 0.25%Ru/CeO₂_700 with a value of 5.3 g_C cm_cat⁻³ h⁻¹. On the other hand, both reference catalysts exhibit the highest efficiency with values of 14.5 and 15.1 g_C cm_cat⁻³ h⁻¹ for 5.00%Ru/CNF and 5.02%Ru/AC, respectively.

Fig. 4 shows the relationship between gas composition and conversion of glycerol feed (3 wt%) achieved at 400 °C and 270 bar over Ru/CeO₂ and Ru/CNF catalysts. In both cases the main reaction products were H₂, CO₂, and CH₄, while CO was marginal. Both classes of catalysts performed very similarly over the range of apparent conversion rates tested. However, Ru/CeO₂ seems to slightly favour CH₄ production, while Ru/CNF slightly favours hydrogen production. For both types of catalysts and at conversions below 90%, the H₂ and CH₄ concentrations in the gas were in the range of 34–46 and 18–28 vol%, respectively. Interestingly, when conversion reaches a level above 90%, a noticeable increase in CH₄ production was observed at the expense of H₂. On the other hand, no influence of glycerol conversion and catalyst type on the amount of CO₂ (approx. 30–40 vol%) and CO (less than 1 vol%) produced was detected. In our previous research,^25,48 slightly higher CH₄ and lower H₂ concentrations were measured in the produced gas when Ru/CNF catalysts were used, which is directly linked to the different feed concentrations.^31,50

Fig. 4 Relationship between gas composition (H₂, CH₄, CO₂ and CO in A, B, C and D, respectively) and conversion of glycerol (3 wt%) achieved at 400 °C and 270 bar over Ru/CeO₂ and Ru/CNF catalysts.

SCWG experiments were also carried out using 10 wt% of glycerol, maintaining the pressure and temperature used previously (Fig. S9). Increasing the glycerol content in the feed resulted in a significant change in conversion, which significantly dropped with most points below 30 and 50% for both Ru/CeO₂ and Ru/CNF catalysts, respectively. This notably influenced the gas composition, with overall higher H₂ and lower CH₄ concentrations for the Ru/CeO₂ than for the Ru/CNF catalyst. However, there was a clear difference between the two catalysts, with a much higher H₂ concentration, a higher CO₂ and a lower CH₄ concentration for Ru/CeO₂ when compared with Ru/CNF. Surprisingly, CO levels were found to be very different between the two catalysts, with a concentration peaking at 25 vol% for Ru/CNF at 17% conversion, while for Ru/CeO₂ the CO concentration always remained very low. A similar behaviour was found for Ru/CeO₂ catalysts when changing from 3 to 10 wt% glycerol, while a comparison was not possible for Ru/CNF because of the lack of data at low conversion and 3 wt% glycerol. The different gas composition can be explained by an accelerated water gas shift (WGS) reaction over CeO₂, which oxidizes CO with water to produce H₂ and CO₂.

To recognise the influence of the CeO₂ support itself on the glycerol gasification process, experiments were carried out using Ru-free CeO₂_700 (Fig. S10A). This sample (previously calcined at 700 °C) showed a low carbon conversion of 12.2 ± 0.2% at a pressure of 275 bar, a temperature of 427 °C and a low feed flow rate of 3.7 g min⁻¹ of 3 wt % glycerol in water. When the temperature was lowered to 400°C and the feed mass flow rate was doubled (to 7.3 g min⁻¹), the conversion dropped by a factor of three, reaching just 3.2 ± 0.1%. However, this had no significant effect on the gas composition, where the main products in both cases were H₂ (approx. 50 vol%) and CO₂ (approx. 30 vol%). On the other hand, CO was in the range of 14–17 vol%, while CH₄ was only 1 vol%. No other gaseous products were observed. These observations, compared to data in Fig. 4, indicates that there is a synergy between Ru and CeO₂. Indeed, CeO₂ alone produces large amounts of CO but no CH₄ at low conversion, while with Ru/CeO₂, CH₄ formation is observed but no CO. Interestingly, for the same sample, but after an additional reduction step in H₂ under the same conditions as during the pretreatment of Ru/CeO₂ catalyst precursors, a lower carbon-based conversion of glycerol as well as a different composition of the produced gas were observed (Fig. S10B). In this case, the conversion was in the range of 5–8 ± 1%, depending on the SCWG process conditions. In terms of gas composition, apart from minor changes in the concentration of H₂, CO₂, and CH₄, the main difference was the detection of propane instead of CO. The cause of the selective production of propane, through successive dehydration and hydrogeneration, instead of CO cannot be explained with the information gathered. However, it can be stated that the different selectivity of both supports is clearly due to the different thermal pre-treatments of the CeO₂, which likely affected the reduction state of cerium oxide and the concentration of oxygen vacancies,⁵¹ despite an exposure to air before loading the material in the reactor. CeO₂ participates in the SCWG process, but its activity is negligible compared to Ru, and its selectivity depends on the pretreatment of the solid.

Ru/CeO₂ evolution during SCWG

After the modification of the CeO₂ support by incipient wetness impregnation with the Ru precursor and subsequent thermal reduction with H₂ at 300 °C, no change in CeO₂ crystallinity was observed and no other additional phases were detected by XRD (Fig. S11A–D). The lack of a crystalline Ru⁰ or any other ruthenium-based diffraction lines in the XRD patterns for all Ru/CeO₂ samples might be a result of the very low loading and/or the low size of Ru particles (less than a few nanometers). In the case of both reference catalysts, which contained a higher amount of Ru (5 wt%), also no Ru phase signals were observed, which may be related to the much higher porosity of these catalysts (Table 4). Since only a small amount of Ru was deposited on the CeO₂ surface, the catalysts showed a porosity comparable to the unmodified CeO₂ supports (Table 4). The 0.11%Ru/CeO₂_900 catalyst exhibited low S_BET and V_total values (3.0 m² g⁻¹ and 0.023 cm³ g⁻¹, respectively), because of the low porosity of the support, calcined at 900 °C. All Ru/CeO₂ catalysts containing higher Ru loadings showed values close to the values for their support CeO₂_700. After the glycerol gasification experiments, the XRD patterns recorded for the Ru/CeO₂ samples remained unchanged and no characteristic reflections of the CeCO₃OH phase were observed in any of the tests performed in the continuous setup (Fig. S11). Moreover, the average CeO₂ crystallite size evaluated via Rietveld refinement increased slightly for the spent catalysts (Table 4). For all higher Ru loading catalysts (0.25–1.00 wt%), the V_total and S_BET values decreased by about 25–35% and 50–60%, respectively, after SCWG tests. These changes suggest partial degradation of the catalysts, probably due to coarsening of CeO₂ (“sintering”) and/or blocking of existing pores, e.g. by recrystallization under SCW/SCWG conditions. In turn, such a reduction of both parameters was neither observed for 0.11%Ru/CeO₂_900 nor for CeO₂_900. For both reference catalysts, an increase in S_BET and V_total values was observed after the SCWG experiment, which is consistent with the literature²⁰ and is related to the leaching and/or gasification of the least stable carbon present in the supports.

Table 4 Evolution of specific surface area, pore volume and crystallite size of CeO₂ in Ru/CeO₂ catalysts before and after (marked “_S”) continuous SCWG of glycerol

Sample	SBET [m^2 g^−1]	Vtotal [cm^3 g^−1]	Crystallite size^a [nm]
a Crystallite size of CeO2 evaluated via Rietveld refinement. See Table S4 for comparison with data from the Scherrer equation.b Test performed in batch experiment with pure water for 1 h at 270 bar, 430 °C, to mimic the conditions the catalyst is exposed to prior to the start of a continuous test; n/a = not applicable.
CeO2_900	3.0	0.022	82
0.11%Ru/CeO2_900	3.0	0.023	91
0.11%Ru/CeO2_900_S	2.9	0.022	94
CeO2_700	30	0.068	31
CeO2_700_S	18	0.067	29
0.25%Ru/CeO2_700	28	0.064	30
0.25%Ru/CeO2_700_S	11	0.042	36
0.25%Ru/CeO2_700_SCW^b	16	0.065	34
0.44%Ru/CeO2_700	34	0.078	26
0.44%Ru/CeO2_700_S	17	0.060	30
1.00%Ru/CeO2_700	31	0.068	26
1.00%Ru/CeO2_700_S	12	0.045	26
5.00%Ru/CNF	255	1.93	n/a.
5.00%Ru/CNF_S	292	2.15	n/a
5.02%Ru/AC	796	0.30	n/a
5.02%Ru/AC_S	837	0.32	n/a

---

Importantly, these changes did not result in a loss of activity during the gasification process (Fig. S7 and S8), which indicate either that they have no impact in the total surface of active ruthenium, or that such loss occurred before the SCWG experiments. To confirm this hypothesis, a stability test was performed in only SCW to study its effect on surface area and pore volume. Indeed, already after 1 h under SCW conditions (275 bar, 430 °C), the BET specific surface area of 0.25% Ru/CeO₂_700 decreased from 28 to 16 m² g⁻¹, while the total pore volume remained almost unchanged (0.064 cm³ g⁻¹ before and 0.065 cm³ g⁻¹ after experiment). This result confirms that the loss of surface area of the catalyst during the test mostly occurs during the heating up and stabilisation phase in pure water flow. Note that there is a lack of consistency between the crystallite size and the BET surface area, for example between 1.00%Ru/CeO₂_700 and 1.00%Ru/CeO₂_700_S for which S_BET decreased threefold but the crystallite size did not change (Table 4 and S4). This can be explained by coarsening leading do a decrease of the intercrystallite voids (aggregation of crystals to a large pore-free particle), without crystallite coalescence (two crystals fusing to a larger one). Another explanation would be the filling of intercrystallite voids with amorphous material which do not present diffuse scattering peaks in XRD.

The incipient wetness impregnation technique enables the deposition of very fine Ru particles on the support surface, as we have shown in our previous study on Ru/CNF.²⁵ In the case of this class of materials, Ru nanoparticles could be clearly identified, because of the high contrast compared to the CNF matrix. However, in the case of the discussed Ru/CeO₂ samples, the very low contrast between Ru nanoparticles (Ru⁰ core with a RuO₂ shell),⁵² makes the evaluation of Ru particle size distribution unrealistic by electron microscopy, as EDX analysis would be necessary to confirm the composition of each observed nanoparticles. TEM revealed first highly crystalline CeO₂ particles with a size in the same range as listed in Table 4, and with a low-porosity consisting of cavities in the 1–5 nm diameter range. As it can be seen in Fig. 5, pure CeO₂ showed 1–4 nm particles that could be observed at the edge of CeO₂ crystals on 0.44%Ru/CeO₂_700_S, as well as on Ru-free CeO₂_700_S. STEM-EDX analysis on several particles proved that most were CeO₂, but also a few were Ru nanoparticles. Additionally, the occurrence of these darker spots (potential Ru nanoparticles) is highly dependent on the appropriate focusing during the TEM measurement (Fig. S12). Therefore, for Ru/CeO₂ (regardless of the Ru loading) it is not possible to precisely identify Ru particles, and consequently their size and dispersion cannot be determined in a statistically meaningful way. Nevertheless, it should be emphasized that the identified particles (which can be assigned to both Ru- and CeO₂-based species) in the Ru/CeO₂ systems are in the range of 1.0 ± 0.2 to 1.6 ± 0.4 nm (Table S5). As a result of glycerol gasification, the size of these particles increased to values of 1.8 ± 0.5–2.5 ± 0.5 nm. A similar increase in Ru particle size after SCWG experiments was also observed for the reference 5.00% Ru/CNF catalyst used in this study (from 2.0 ± 0.6 to 3.1 ± 0.8 nm), and is also consistent with our previous findings, where stabilization in the range of 2.2–2.8 nm was observed in Ru/CNF catalysts (Ru loading range 1–30 wt%).^20,25 Interestingly, after 1 h of exposure of the 0.25%Ru/CeO₂_700 sample solely to SCW conditions (270 bar, 430 °C), the identified particles (which still can be assigned to both Ru- and CeO₂-based species) increased only slightly from 1.6 ± 0.4 to 1.7 ± 0.4 nm. It should be emphasized that the same observation (of lack of Ru particle size growth in pure SCW) was also confirmed in the case of Ru/CNF.²⁵

Fig. 5 HR-TEM images of the CeO₂_700_S support (A) and 0.44%Ru/CeO₂_700_S catalyst (B).

In our previous study with Ru/CNF, no correlation was found between Ru particle size and catalysts activity, however, a clear correlation was observed between the TOF and the surface density of active ruthenium on the catalyst surface.²⁵ Since the Ru dispersion could not be determined and thus the TOF value not calculated reliably in this study, the catalyst activity was expressed as the apparent carbon conversion rate as a function of total Ru surface density (Fig. 6). In this comparison, Ru/CeO₂ catalysts with 0.11–0.44 wt% Ru were the most efficient, achieving high apparent conversion rates (128–233 mol_C mol_Ru⁻¹ min⁻¹), with a total Ru surface density of 1.4–2.3 atom_Ru nm⁻² (maximum apparent conversion rate reached at 1.6 atom_Ru nm⁻²). In turn, for 1.00%Ru/CeO₂_700, the lowest conversion factor was recorded, which was only 59 mol_C mol_Ru⁻¹ min⁻¹, which was accompanied by a high value for the total Ru density equal to 4.9 atom_Ru nm⁻². Interestingly, both reference carbon-supported catalysts also follow this trend, as high activity was achieved for 5.00%Ru/CNF (181 mol_C mol_Ru⁻¹ min⁻¹ at 1.0 atom_Ru nm⁻²), while 5.02%Ru/AC was characterized by a much lower efficiency (69 mol_C mol_Ru⁻¹ min⁻¹), which was related to the high Ru density (7.2 atom_Ru nm⁻²).

Fig. 6 Apparent carbon conversion rate over Ru/CeO₂, Ru/CNF and Ru/AC catalysts as a function of the total surface density of Ru. The asterisks (*) indicate a possible underestimation of the apparent conversion rate because the conversion was not low enough (<60%).

While the exact origin of the performance decrease with increasing Ru loading remains unclear, two plausible explanations can be proposed: 1. an optimal balance between Ru active sites and the carbon support surface might exist, as the carbon support itself has demonstrated intrinsic activity in the hydrothermal conversion of organic compounds;^20,53 2. for Ru particles supported on carbon a mild particle growth was reported under reductive supercritical water conditions,²⁵ and a similar particle size growth can be expected with CeO₂ support. Although the mechanism responsible for this growth is not fully understood, atomic migration–coalescence (Ostwald ripening)—via detachment and surface diffusion of Ru atoms on the support—may be involved. At higher Ru loadings, the increased surface density of Ru atoms would statistically reduce the fraction of isolated or low coordination Ru surface species. As these sites have been suggested to exhibit particularly high intrinsic activity,²⁵ an increase in Ru loading could therefore lead to a lower overall catalytic efficiency due to a decreased proportion of highly active sites.

Literature reports consistently show that CeO₂ exhibits negligible acidity and only very weak basicity associated with surface hydroxyl groups.^54,55 CO₂-TPD and NH₃-TPD profiles of CeO₂ and Ru/CeO₂ remain essentially unchanged upon Ru addition, indicating that variations in Ru loading do not affect the acid–base characteristics of the support, particularly at loadings below 1 wt%. Therefore, differences in catalytic performance with Ru loading cannot be attributed to changes in acidity or basicity of CeO₂.

Ru/CeO₂ evolution during SCWG

We have previously shown that the loss of activity of Ru/AC catalysts during SCWG is not related to leaching of Ru,⁵⁶ while catalyst coking as well as the decrease of dispersion of the active phase may be the main mechanism of its rapid deactivation. To determine the stability of the catalysts over time on stream, the 5.00%Ru/CNF and 0.44%Ru/CeO₂_700 samples were selected and subjected to the experiment under constant SCWG conditions (400 °C, 270 bar and WHSV equal to 5400 ± 100 g_glycerol g_Ru⁻¹ h⁻¹ for 5.00% Ru/CNF, and 7200 ± 600 g_glycerol g_Ru⁻¹ h⁻¹ for 0.44% Ru/CeO₂_700). In both cases, a similar decreasing trend in C-based glycerol conversion was observed during the SCWG experiment (Fig. 7). By extrapolation, the decrease in conversion expected after hundreds of hours on stream is around 46 and 49%, respectively (Fig. S13).

Fig. 7 Evolution of carbon (glycerol) conversion over 0.44%Ru/CeO₂_700 and 5.00%Ru/CNF and as a function of time on stream.

One of the factors responsible for the decrease in efficiency of both types of catalysts may be the increase in Ru particle size, which however cannot be reliably determined by TEM/STEM. On the other hand, it is also well known that catalysts undergo coking, which is one of the factors negatively affecting their performance in the SCWG process. Therefore, TGA-MS analysis was performed for all spent Ru/CeO₂ samples to determine the amount of possible coke deposits (Fig. S14). Additionally, fresh catalysts and the corresponding supports (CeO₂_700 and CeO₂_900) were analysed under the same conditions (Fig. S15). Fresh Ru-containing samples (previously reduced at 300 °C) show high thermal stability, as confirmed by the insignificant mass change (usually less than 0.6 wt%, a higher value was noted only for 0.44%Ru/CeO₂_700) recorded in the temperature range from 110 to 1000 °C (Table 5). Comparable mass decreases were also detected for pure CeO₂ samples, confirming that the mass decrease of fresh Ru/CeO₂ catalysts are related to the changes occurring in the CeO₂ matrix, which can be attributed to the desorption of chemically adsorbed oxygen, surface carbonates and/or desorption of surface lattice oxygen atoms.⁵⁷ In turn, a different behaviour was observed for the spent catalysts. For the 0.25%Ru/CeO₂_700_S and 1.00%Ru/CeO₂_700_S samples, the largest mass loss was observed with 4.01 and 3.83 wt%, respectively, with a sharp decrease at a temperature of about 300 °C. This was accompanied by the release of a large amount of CO₂ (mass line m/z = 44), which clearly proves that at this temperature the coke deposited on the catalysts during the SCWG experiment was oxidized. If one considers the temperature range up to 400 °C (CO₂ release occurs below this temperature) as well as the slight mass changes in the corresponding fresh catalysts, 33.7 and 30.8 mg coke per 1 g of spent catalyst (mg_C g_cat⁻¹) can be estimated, respectively. Much less coke was found on 0.44%Ru/CeO₂_700_S, which amounted to only 2.3 mg_C g_cat⁻¹ corresponding to 0.23 wt% loss. In turn, the spent sample containing the lowest amount of Ru (0.11%Ru/CeO₂_900_S) did not show any mass change, which indicates negligible carbon deposits on this sample. Insignificant carbon deposits were also measured for the spent Ru-free CeO₂ sample (CeO₂_700_S). The comparison does not reveal a clear correlation between Ru loading and the extent of coking, and the reasons why certain samples exhibit higher coke deposition than others remain unclear.

Table 5 Mass loss during TGA measurements and determined amount of carbon deposit in Ru/CeO₂ samples. n.d. = not detected, due to the difference between fresh and spent materials

Sample	Mass loss [wt%]		Coke deposit after SCWG^a [mgC gcat^−1]
	110–400 °C	110–1000 °C
a Based on the mass difference between the fresh and spent sample in the range of 110–400 °C.b These samples were analysed twice with consistent results.
CeO2_700	0.27	0.41	—
CeO2_700_S	0.26	0.43	—
CeO2_900	0.04	0.07	—
0.11%Ru/CeO2_900	0.04	0.10	—
0.11%Ru/CeO2_900_S^b	0.03	0.09	n.d.
0.25%Ru/CeO2_700	0.25	0.42	—
0.25%Ru/CeO2_700_S	3.62	4.01	33.7
0.44%Ru/CeO2_700	0.63	0.93	—
0.44%Ru/CeO2_700_S^b	0.86	1.16	2.3
1.00%Ru/CeO2_700	0.36	0.56	—
1.00%Ru/CeO2_700_S	3.45	3.83	30.8

---

In the case of carbon-based catalysts (e.g. Ru/AC, Ru/CNF), it is not possible to efficiently regenerate them by simple selective oxidation of the carbon deposits at elevated temperature, which is one of the main disadvantages of such systems. Conversely, this regeneration strategy can be readily implemented for Ru/CeO₂ systems, which represents a clear advantage of these catalysts. Additionally, in this work we have evaluated the possibility of catalyst bed regeneration using H₂O₂ as a mild oxidant at a moderate temperature of 110 °C, already applied to regenerate Ru/AC catalyst poisoned by sulfur.²⁶ A spent sample containing 0.44 wt% Ru was selected and treated with 5 wt% H₂O₂ solution for this study. The collected results presented in Fig. 8 clearly confirm the efficiency of the proposed method to remove coke, which neither showed any weight loss associated with carbon residue oxidation at temperature around 300 °C nor related CO₂ emission in the regenerated sample (0.44%Ru/CeO₂_700_R). It is worth emphasizing that the proposed regeneration technique can be used directly in the high-pressure reactor, without the need to remove the catalyst. Moreover, such treatment does not negatively affect the physicochemical properties of the recovered sample which is confirmed by XRD technique and N₂ adsorption.

Fig. 8 TGA profiles (black line) and m/z = 44 MS line assigned to CO₂ (blue line) for spent 0.44%Ru/CeO₂_700_S catalyst (A) and after regeneration with H₂O₂ solution at 110°C (B).

Conclusions

Overall, this study shows that CeO₂ is a promising non-carbon support for Ru in cSCWG, combining high activity, stability under optimized conditions, and practical regenerability – features that make it a strong candidate for industrial-scale biomass conversion.

CeO₂ undergoes transformation to CeCO₃OH under hydrothermal batch conditions, causing surface area loss, but remains stable in continuous SCW provided pH > 6 and CO₂ exposure below 400 °C is avoided. Stability can be improved by pre-coarsening, with optimal calcination between 700–900 °C. Ru/CeO₂ catalysts exhibited excellent performance in cSCWG of glycerol at 400 °C and 270 bar, reaching conversion rates up to three times higher than commercial Ru/AC while showing similar selectivity trends. An optimal Ru surface loading of 1.5–2.0 atom_Ru nm⁻² was identified, slightly higher than on CNF-supported catalysts. A key advantage is the ease of coke and sulfur removal via mild oxidative treatments (O₂/air or H₂O₂), enabling efficient regeneration and lowering process costs. Future work should focus on elucidating the influence of catalyst preparation on coking behaviour and on optimizing regeneration protocols to ensure complete recovery of catalytic performance. From an industrial perspective, higher Ru loading is preferred as it reduces CAPEX linked to the costly volume of high-pressure, high-temperature reactors. However, this should not be at the cost of a much lower intrinsic activity of costly ruthenium. This underlines the importance of developing CeO₂-supported catalysts with stable and high surface area, enabling high Ru surface loading and improved volumetric catalytic performance for cSCWG. This could be reached, for example, by doping CeO₂, e.g. with Zr or Y, to improve the hydrothermal stability of the high surface area inorganic oxide.

Author contributions

T. Kondratowicz: conceptualization, investigation, visualization, writing – original draft, writing – review and editing; A. Quenet: conceptualization, investigation, visualization, writing – original draft, writing – review and editing; X. Li: investigation, writing – review and editing; A. Testino: writing – review and editing; O. Kröcher: supervision, writing – review and editing; F. Vogel: conceptualization, funding acquisition, writing – review and editing; D. Baudouin: supervision, conceptualization, funding acquisition, visualization, writing – review and editing.

Conflicts of interest

A. Quenet, F. Vogel and D. Baudouin declare that a patent application has been filed related to the technology described in this manuscript.

Data availability

Data for this article, including raw experimental data and raw analytical files (TGA, N₂ adsorption, XRD, TEM) are available on Zenodo. See DOI: https://doi.org/10.5281/zenodo.20398121.

Supplementary information (SI) is available. See DOI: https://doi.org/10.1039/d5cy01486g.

Acknowledgements

The authors acknowledge the financial support of BIOCTANE project with Grant No. 101084336 funded by the European Union. The authors would also like to thank Canio Scarfiello for the laboratory work on CeO₂ stability, Inge Stockinger as well as Patricia Pizarro de Oro and Sinay Sibaneth Turriziani (Rey Juan Carlos University) for performing the ICP-OES measurements, Songlan Sun for helping with the regeneration of the spent catalyst, Elisabeth Agnes Müller Gubler, Ivo Alxneit and Emiliya Poghosyan for their support with electron microscopy, Amel Tumbul, Norbert Schmid and Thomas Brunner for their support with the continuous test rig Konti-I.

Notes and references

1. K. L. Ebi, J. Vanos, J. W. Baldwin, J. E. Bell, D. M. Hondula, N. A. Errett, K. Hayes, C. E. Reid, S. Saha, J. Spector and P. Berry, Annu. Rev. Public Health, 2021, 42, 293–315.
2. K. Abbass, M. Z. Qasim, H. Song, M. Murshed, H. Mahmood and I. Younis, Environ. Sci. Pollut. Res., 2022, 29, 42539–42559.
3. The Paris Agreement | UNFCCC, https://unfccc.int/process-and-meetings/the-paris-agreement, (accessed May 9, 2025).
4. F. Gale, D. Goodwin, H. Lovell, H. Murphy-Gregory, K. Beasy and M. Schoen, Int. J. Hydrogen Energy, 2024, 59, 654–667.
5. M. R. Errera, T. A. da C. Dias, D. M. Y. Maya and E. E. S. Lora, Biomass Bioenergy, 2023, 170, 106721.
6. W. de Jong and J. R. van Ommen, in Biomass as a sustainable energy source for the future: fundamentals of conversion processes, ed. W. de Jong and J. R. van Ommen, Wiley, Hoboken, New Jersey, 2014, ch. III.
7. J. Giuntoli, PhD thesis, TU Delft, 2010.
8. A. A. Peterson, F. Vogel, R. P. Lachance, M. Froling, M. J. Antal and J. W. Tester, Energy Environ. Sci., 2008, 1, 32.
9. F. Vogel, in Encyclopedia of Sustainability Science and Technology, ed. R. Meyers, Springer, New York, 2017, pp. 1251–1295.
10. F. Vogel, Volume 2: Heterogeneous Catalysis, in Handbook of Green Chemistry, ed. R. H. Crabtree, Wiley, Weinheim, 2010, pp. 281–324.
11. N. Boukis, E. Hauer, S. Herbig, J. Sauer and F. Vogel, Biomass Convers. Biorefin., 2017, 7, 415–424.
12. D. C. Elliott, Biofuels, Bioprod. Biorefin., 2008, 2, 254–265.
13. B. Pinkard, J. Kramlich, P. Reinhall and I. Novosselov, in Advanced Supercritical Fluids Technologies, ed. I. Pioro, IntechOpen, 2019.
14. R. Muhlke, J.-B. Castaing, M. Morel and V. Donat, Hydrothermal gasification - White paper, GRTgaz, Bois-Colombes, France, 2023.
15. M. Schubert, J. W. Regler and F. Vogel, J. Supercrit. Fluids, 2010, 52, 99–112.
16. P. A. Marrone, M. Hodes, K. A. Smith and J. W. Tester, J. Supercrit. Fluids, 2004, 29, 289–312.
17. J. Reimer, G. Peng, S. Viereck, E. De Boni, J. Breinl and F. Vogel, J. Supercrit. Fluids, 2016, 117, 113–121.
18. G. J. DiLeo and P. E. Savage, J. Supercrit. Fluids, 2006, 39, 228–232.
19. E. Galiwango, J. Butler and S. Lotfi, Fuels, 2024, 5, 375–393.
20. C. M. Hunston, PhD thesis, EPFL, 2021.
21. J. N. Jocz, P. E. Savage and L. T. Thompson, Ind. Eng. Chem. Res., 2018, 57, 8655–8663.
22. J. N. Jocz, L. T. Thompson and P. E. Savage, Chem. Mater., 2018, 30, 1218–1229.
23. S. Li, P. E. Savage and L. Guo, J. Supercrit. Fluids, 2018, 135, 188–197.
24. J. G. Dickinson and P. E. Savage, J. Mol. Catal. A: Chem., 2014, 388–389, 56–65.
25. C. Hunston, D. Baudouin, L. Koning, A. Agarwal, O. Kröcher and F. Vogel, Appl. Catal., A, 2023, 320, 121956.
26. M. Dreher, M. Steib, M. Nachtegaal, J. Wambach and F. Vogel, ChemCatChem, 2014, 6, 626–633.
27. M. Dreher, B. Johnson, A. A. Peterson, M. Nachtegaal, J. Wambach and F. Vogel, J. Catal., 2013, 301, 38–45.
28. M. H. Waldner, PhD thesis, ETH Zurich, 2007.
29. TreaTech, personal communication, May 2025.
30. G. Peng, C. Ludwig and F. Vogel, ChemCatChem, 2016, 8, 139–141.
31. H. Zöhrer, F. Mayr and F. Vogel, Energy Fuels, 2013, 27, 4739–4747.
32. A. Abdulrasheed, A. A. Jalil, Y. Gambo, M. Ibrahim, H. U. Hambali and M. Y. Shahul Hamid, Renewable Sustainable Energy Rev., 2019, 108, 175–193.
33. P. Li, X. Chen, Y. Li and J. W. Schwank, Catal. Today, 2019, 327, 90–115.
34. E. Aneggi, C. de Leitenburg, G. Dolcetti and A. Trovarelli, Catal. Today, 2006, 114, 40–47.
35. C. de Leitenburg, A. Trovarelli, J. Llorca, F. Cavani and G. Bini, Appl. Catal., A, 1996, 139, 161–173.
36. J. Liu, Z. Zhao, C. Xu and J. Liu, Chin. J. Catal., 2019, 40, 1438–1487.
37. K. Yuan, X.-C. Sun, H.-J. Yin, L. Zhou, H.-C. Liu, C.-H. Yan and Y.-W. Zhang, J. Energy Chem., 2022, 67, 241–249.
38. B.-S. Cheon, H.-M. Kim, J.-H. Hwang and D.-W. Jeong, Fuel, 2025, 383, 133777.
39. C. Alvarez-Galvan, P. G. Lustemberg, F. E. Oropeza, B. Bachiller-Baeza, M. Dapena Ospina, M. Herranz, J. Cebollada, L. Collado, J. M. Campos-Martin, V. A. de la Peña-O'Shea, J. A. Alonso and M. V. Ganduglia-Pirovano, ACS Appl. Mater. Interfaces, 2022, 14, 50739–50750.
40. R. Kopelent, A. Tereshchenko, A. Guda, G. Smolentsev, L. Artiglia, V. L. Sushkevich, A. Bugaev, I. I. Sadykov, T. Baidya, M. Bodnarchuk, J. A. van Bokhoven, M. Nachtegaal and O. V. Safonova, ACS Catal., 2021, 11, 9435–9449.
41. A. Solís-García, K. Portillo-Cortez, D. Domínguez, S. Fuentes-Moyado, J. N. Díaz de León, T. A. Zepeda and U. Caudillo-Flores, Catalysts, 2025, 15, 301.
42. F. Cao, Y. Xiao, Z. Zhang, J. Li, Z. Xia, X. Hu, Y. Ma and Y. Qu, J. Catal., 2022, 414, 25–32.
43. D. Baudouin, H. Xiang and F. Vogel, Biomass Bioenergy, 2022, 164, 106529.
44. Q. Guan, X. Huang, J. Liu, J. Gu, R. Miao, Q. Chen and P. Ning, Chem. Eng. J., 2016, 283, 358–365.
45. L. He, Y. Ren, Y. Fu, B. Yue, S. C. E. Tsang and H. He, Molecules, 2019, 24, 526.
46. K. Michiba, R. Miyawaki, T. Minakawa, Y. Terada, I. Nakai and S. Matsubara, J. Mineral. Petrol. Sci., 2013, 108(6), 326–334.
47. X. Mao, X. Xia, J. Li, C. Chen, X. Gu, S. Li and Y.-P. Lan, J. Alloys Compd., 2021, 870, 159424.
48. D. Salionov, C. Hunston, F. Vogel, D. Baudouin and S. Bjelić, J. Catal., 2023, 426, 257–269.
49. F. Vogel, T. Brunner, D. Baudouin, R. Zambrano, G. Peng, A. Mian and M. Brandenberger, Pilotanlage zur katalytischen hydrothermalen Vergasung nasser Biomasse - HydroPilot Pilot plant for the catalytic hydrothermal gasification of wet biomass, SFOE, Villigen, Switzerland, 2023.
50. D. J. M. de Vlieger, A. G. Chakinala, L. Lefferts, S. R. A. Kersten, K. Seshan and D. W. F. Brilman, Appl. Catal., A, 2012, 111–112, 536–544.
51. G. Feng, W. Han, Z. Wang, F. Li and W. Xue, Catalysts, 2018, 8, 535.
52. R. Coloma Ribera, R. W. E. van de Kruijs, S. Kokke, E. Zoethout, A. E. Yakshin and F. Bijkerk, Appl. Phys. Lett., 2014, 105, 131601.
53. Y. Matsumura, Y. Suganuma, T. Ichikawa, W. Kim, Y. Nakashimada and K. Nishida, ACS Omega, 2021, 6(36), 23442–23446.
54. Y. Zhao, C. Xi, S. Gao, Y. Wang, H. Wang, P. Sun and Z. Wu, RSC Adv., 2023, 13, 7037–7044.
55. N. Elia, J. Estephane, C. Poupin, B. El Khoury, L. Pirault-Roy, S. Aouad and E. A. Aad, ChemCatChem, 2021, 13, 1559–1567.
56. C. Hunston, D. Baudouin, M. Tarik, O. Kröcher and F. Vogel, Catal. Sci. Technol., 2021, 11, 7431–7444.
57. N. H. T. Phan, C. C. Nguyen and M. T. N. Dinh, RSC Adv., 2023, 13, 13354–13364.

---

Footnote

† These authors contributed equally to the research and are to be considered co first authors.

---