Catalytic Oxidation of Ethanol for Treatment of Commercially Produced Carbon Dioxide Using Aqueous Sol–Gel-Derived Catalyst Materials

Abstract

The purpose of this study was to design a Pt/TiO₂–ZrO₂ catalytic-based treatment system to remove ethanol and oxygen (O₂) from a gaseous feed stream. The ultimate target application was the conversion of ethanol and O₂ to carbon dioxide (CO₂) and water (H₂O) from a feed stream of CO₂ in a commercial beer brewing operation. Bench-scale reactions were performed at 250 °C and 300 °C, representing two temperatures under practical consideration for a full-scale catalytic reactor. The target gaseous feed stream would be expected to have a relatively low (near-stoichiometric) concentration of O₂, so the effect of O₂ concentration was also studied. On the bench scale, ethanol was completely converted to CO₂ under low flow rate conditions, and the reactions proceeded through volatile and non-volatile reaction intermediates. Results from the bench-scale tests were used to make predictions for designing a pilot-scale catalytic reactor under conditions of high and low O₂ concentration. A pilot-scale reactor was constructed and installed in a commercial brewing facility, and results from testing the pilot-scale reactor are also presented. The pilot-scale system reduced the feed stream ethanol concentrations by 99.9% while concomitantly reducing the O₂ concentrations over the course of a six-day demonstration period without generating unacceptable levels of byproducts.

1. Introduction

Ethanol (CH₃CH₂OH) is a common chemical that is widely used as an antiseptic, a general solvent, a fuel, and a fuel additive. It is also generated in the fermentation process of making alcoholic beverages like beer, wine, and liquor. Ethanol is a highly volatile chemical; consequently, ethanol vapor is often present at significant levels in areas where pure ethanol liquid or solutions containing ethanol are used. Removal of this gas-phase ethanol can be desirable for pollution abatement (i.e., if the ethanol-containing gas is eventually vented to the atmosphere) or if the goal is to recycle a bulk gas (e.g., carbon dioxide) for commercial use from a feed that contains appreciable amounts of ethanol vapor.

Other researchers have studied the catalytic oxidation of gas-phase ethanol under a variety of conditions [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]. The majority of these papers have focused on evaluating novel catalysts for the conversion of ethanol to carbon dioxide (CO₂), with temperature being the main operational parameter of interest. Temperatures of 200–500 °C have typically been reported for complete mineralization of ethanol to CO₂. Some researchers have focused on the effect of low concentrations of oxygen (O₂); typically, these experiments involved removing O₂ completely to study the role of lattice O₂ on the reaction mechanism [2,3]. Acetaldehyde is the main gas-phase byproduct observed during the catalytic oxidation of ethanol [1,5,6,10,11]; however, acetic acid, ethylene, methane, and ethyl acetate have also been reported [7,9]. The complete mineralization of ethanol to CO₂ and water (H₂O) is given by Equation (1):

CH₃CH₂OH + 3 O₂ → 2 CO₂ + 3 H₂O

(1)

The catalyst used in this study was composed of titanium dioxide (TiO₂) and zirconium dioxide (ZrO₂) that were prepared using a sol–gel processing method and subsequently platinized. The catalyst was coated onto high surface area supports. Others have reported using TiO₂–ZrO₂ materials for catalytic purposes. In 2005, Reddy and Khan [18] published the first comprehensive review that summarized TiO₂–ZrO₂ synthesis methods, characterization, and applications as catalysts and photocatalysts and supports for the purpose of facilitating different types of reactions. The mixed oxide has been demonstrated to have enhanced performance over single oxide materials, resulting from high surface area, high thermal stability, high mechanical strength, and enhanced surface acid–base properties. TiO₂–ZrO₂ materials have been used for partial oxidation, deep oxidation, hydrogenation, hydrotreating, organic transformations, NOx abatement, and photocatalysis. The addition of metals like platinum (Pt) or silver (Ag) further improves the catalytic properties of the material. More recently, TiO₂–ZrO₂ materials have been reported for low-temperature catalysis [19], photocatalysis [20], catalytic dehydrogenation [21], catalytic acetalization, and ketalization [22], and also for anticorrosion coatings [23].

The purpose of this study was to design a catalytic-based treatment system to remove ethanol and O₂ from a gaseous feed stream. The ultimate target application was the conversion of ethanol and O₂ to CO₂ and H₂O from a feed stream of CO₂ in a commercial beer brewing operation. The two main products of beer fermentation are gaseous CO₂ and ethanol. Most of this ethanol is present in solution; however, since ethanol has a high volatility, it is also present in the gaseous CO₂ that is produced. While other volatile organic compounds (VOCs) would be present in such a gaseous feed stream, ethanol would be present at a much higher relative concentration than any other VOC. There are a limited number of studies to date that report the catalytic removal of unwanted compounds from feed streams of CO₂ [24,25].

Due to the presence of undesirable constituents like air (i.e., N₂ and O₂) and VOCs (e.g., ethanol), many breweries discharge this CO₂-rich gaseous “waste” stream to the ambient air, which has a negative environmental impact. The technology described herein considers the treatment and reuse of this CO₂ stream, rendering it a “green technology” that could be used by commercial brewing operations. Craft breweries (as well as major breweries) are becoming increasingly environmentally conscious. Many of these companies actively promote their dedication to sustainability. The environmental benefits of the technology presented in this manuscript would appeal to these and other like-minded breweries. In addition, it is reasonable to assume that this technology could also be utilized by non-beverage ethanol fermentation plants to treat similar CO₂ waste streams. In a review article, Xu et al. [26] discussed multiple uses for purified CO₂ from ethanol fermentation facilities.

The hypothesis tested in the present study is that under appropriate reactor conditions (flow rate, temperature, and pressure), a catalyst material can be demonstrated to purify a feed stream containing predominantly CO₂ with trace levels of ethanol (300–500 ppmv) and O₂. Bench-scale reactions were performed at 250 °C and 300 °C, representing two temperatures under practical consideration for a full-scale catalytic reactor. The target gaseous feed stream would be expected to have a relatively low (near or sub-stoichiometric) concentration of O₂, so the effect of O₂ concentration was also studied. Results from the bench-scale tests were used to make predictions for designing a pilot-scale catalytic reactor under conditions of high and low O₂ concentration. A pilot-scale reactor was constructed and installed in a commercial brewing facility, and results from testing the pilot-scale reactor are also presented.

2. Results and Discussion

2.1. Ethanol Degradation in Air

Bench-scale experiments were initially conducted with an ethanol mixture that contained 440 ppmv ethanol with the balance pure air (i.e., 21% O₂) using 1.9 g of supported catalyst. The volumetric flow rate of the ethanol mixture was varied to allow for the determination of rate constants. The effluent was monitored for gaseous hydrocarbons and CO₂ at 250 °C and 300 °C.

Figure 1 shows the result of varying the W/F_v ratio (i.e., varying the volumetric flow rate (F_v) with a constant mass (W) of catalyst) on the catalytic performance at 250 °C. For purposes of comparison, the concentrations of all species are expressed as ppmv carbon. For example, the observed concentration of ethanol was multiplied by a factor of two prior to plotting on Figure 1 since there are two carbon atoms for every molecule of ethanol. As shown, the concentration of ethanol decreased rapidly with increasing W/F_v ratio, followed by an increase and subsequent decrease in the concentration of intermediates, followed by an increase in the concentration of CO₂. The only gas-phase intermediate that was detected in the effluent at low W/F_v ratios was acetaldehyde. In the models, acetaldehyde was not treated separately from the other general intermediates because it was not detected in enough of the samples to allow for estimation of a separate rate constant.

The uncorrected data were fitted with a linearized first-order model, as described in the Materials and Methods section (Equation (3)). The linearized first-order model matched the experimental data (better than other models tested: 1/2 order, 3/4 order, 3/2 order, and second order), and the calculated first-order rate constants at 250 °C are as follows: 4.4 × 10⁻³ L g⁻¹ s⁻¹ (ethanol) and 5.7 × 10⁻⁴ L g⁻¹ s⁻¹ (intermediates). These rate constants were then used to calculate predicted concentrations for ethanol, intermediates, and CO₂ as described in the Materials and Methods (Section 3.5, Equations (7)–(9)). The predicted concentrations are shown as solid lines in Figure 1. As can be seen, the first-order model provides an excellent fit to the experimental data. Gas concentration data used to perform the kinetic analyses from the benchtop studies are provided as Supplementary Materials (Table S1 and Figure S1).

The experiments were repeated at a second reactor temperature of 300 °C. The flow rate of the ethanol mixture was again varied to allow for the determination of rate constants, and the effluent was monitored for gaseous hydrocarbons and CO₂. As before, acetaldehyde was the only gas-phase intermediate detected in experiments conducted at low W/F_v ratios. The results of these experiments are shown in Figure 2. The data were again linearized to calculate rate constants, reported as follows: 8.2 × 10⁻³ L g⁻¹ s⁻¹ (ethanol) and 1.3 × 10⁻³ L g⁻¹ s⁻¹ (intermediates). The rate constants were again used to calculate predicted concentrations for ethanol, intermediates, and CO₂, and the predicted concentrations are shown as solid lines in Figure 2. Again, the first-order model provided an excellent fit to the experimental data. As expected, the higher temperature resulted in significantly faster reactions, as evidenced by the larger reaction rate constants at 300 °C (increased by a factor of 1.9 for ethanol and 2.3 for the intermediates). In addition, a smaller W/F_v ratio was required for complete conversion to CO₂. Complete conversion to CO₂ at 300 °C was achieved at a W/F_v ratio that was approximately half of what was required at 250 °C (i.e., ~3000 g s L⁻¹ vs. ~6000 g s L⁻¹, respectively). Gas concentration data used to perform the kinetic analyses from the benchtop studies are provided in the Supplementary Materials (Table S2 and Figure S2).

2.2. Ethanol Degradation in Low Oxygen

As mentioned above, the target application for this study was for the degradation of ethanol in a feed stream containing a low concentration of O₂. The experiments above were repeated; however, this time, the ethanol mixture contained 360 ppmv ethanol, with 1422 ppmv O₂ and the balance nitrogen. For these experiments, a mass of 3.8 g of supported catalyst was used. The volumetric flow rate of the ethanol mixture was again varied at catalyst temperatures of 250 °C and 300 °C to allow for the determination of rate constants, and the effluent was monitored for gaseous hydrocarbons and CO₂. The experimental data were again linearized to calculate rate constants, and the calculated first-order rate constants were 2.7 × 10⁻³ L g⁻¹ s⁻¹ (ethanol at 250 °C), 1.0 × 10⁻⁴ L g⁻¹ s⁻¹ (intermediates at 250 °C), 6.5 × 10⁻³ L g⁻¹ s⁻¹ (ethanol at 300 °C), and 3.3 × 10⁻⁴ L g⁻¹ s⁻¹ (intermediates at 300 °C). Technically, these would be considered “lumped” rate constants that incorporate the k for ethanol, as well as the k for O₂; however, this type of analysis is appropriate for reactor scaling for the intended purpose (i.e., oxidizing ethanol at target O₂ concentration of ~1,500 ppmv). The first-order model again provided a better fit to the experimental data than other models examined. Similar to the experiments conducted in pure air, the reaction rate constants were larger at a temperature of 300 °C as compared to 250 °C (factors of 2.4 and 3.3 increase for ethanol and intermediates, respectively).

In addition to acetaldehyde, a number of other gas-phase intermediates were detected (as additional GC-FID peaks) in experiments conducted at low W/F_v ratios under low O₂ conditions. One of these compounds was identified as ethylene. Additional intermediates were detected as very small peaks in the GC-FID chromatogram. These intermediates had concentrations that were too low to allow for positive identification with the GC-MSD. All gas-phase intermediates were eliminated (i.e., below analytical detection limits) when the total conversion to CO₂ was 85% or higher for experiments conducted at 250 °C or 300 °C.

By comparing the calculated reaction rate constants given above, one observes that O₂ levels had a significant effect. The “lumped” reaction rate constants observed under low O₂ conditions are much lower than the corresponding reaction rate constants in pure air. Consequently, a much larger reactor (i.e., larger W/F_v ratio) would be required to perform these reactions under low O₂ conditions. For ethanol, the reaction rate constant in pure air was larger by a factor of 1.6 at 250 °C and 1.3 at 300 °C. For intermediates, the reaction rate constant in pure air was larger by a factor of 5.7 at 250 °C and 3.9 at 300 °C. These results suggest that the O₂ concentration affected the conversion of intermediates to CO₂ more than the initial ethanol degradation step. Oxygen concentration also strongly affected the gas-phase byproducts observed. With experiments performed in pure air, the only gas-phase intermediate observed was acetaldehyde. With the experiments performed under low O₂, several additional intermediates were observed in the GC-FID chromatograms. As mentioned, these gas-phase intermediates were completely eliminated when the total conversion to CO₂ was 85% or higher.

A final set of experiments was conducted to illustrate the effect of bleeding pure air into a low O₂, ethanol feed stream. The results presented above suggest that this would be a relatively simple, economical option for increasing the rates of reaction. For these experiments, the ethanol cylinder contained 503 ppmv ethanol and 1465 ppmv O₂, with the balance nitrogen. The catalyst (3.8 g) was heated to 250 °C, and the flow rate of the low O₂, ethanol gas mixture was set at 135 mL min⁻¹ for all experiments. To this feed stream, pure air was added at flow rates that varied between 0 and 15 mL min⁻¹. The percent conversion of ethanol to CO₂ was quite low initially (only 10%). However, if pure air was added at approximately 1.5% of the total flow rate (corresponding to ~4500 ppmv O₂), the percent conversion of ethanol to CO₂ increased to 85%. Insufficient data were collected in these experiments to allow for the use of rigorous kinetic models that included the O₂ concentration; however, these results suggest that at ethanol levels of 400–500 ppmv, adding air at a level of 1.5% of the total flow rate (or ~4500 ppmv O₂) is sufficient to increase reaction rates to the levels observed when performing the reactions in air.

2.3. Pilot-Scale Reactor Scaling

The main objective of the bench-scale experiments was to provide data to design a scaled-up unit. The rate constants presented above were used to estimate the size of the reactor (i.e., mass of catalyst) that would be required to treat a feed stream with an ethanol concentration of 400 ppmv and a volumetric flow rate of 100 L min⁻¹. The model was used to predict the required catalyst mass for varying levels of conversion to CO₂; 75, 80, 85, 90, 95, 99, and 99.9% conversions were modeled. The results of this exercise are presented in

Table 1. Predicted masses of catalyst to achieve various levels of complete conversion of ethanol to CO₂ in a pilot-scale reactor ¹.

Conversion to CO2 (%)	Mass Required In Air (kg)		Mass Required in Low O2 (kg)
	250 °C	300 °C	250 °C	300 °C
75	4.5	2.0	24	7.3
80	5.1	2.3	27	8.4
85	6.0	2.7	32	9.8
90	7.1	3.2	39	12
95	9.2	4.1	51	15
99	14	6.1	77	24
99.9	21	9.1	116	35

¹ Ethanol concentration = 400 ppmv (i.e., 800 ppmv carbon); volumetric flow rate = 100 L min⁻¹.

. Depending on the level of conversion desired, the mass of catalyst can be determined. For example, if 85% conversion to CO₂ is desired, the model predicted that 6.0 kg of catalyst would be required if the reaction were performed in air at 250 °C. However, if the reactor temperature was increased to 300 °C in air, only 2.7 kg of catalyst would be needed. If the reaction was conducted under conditions of low O₂, the required masses of catalyst would be 32 kg and 9.8 kg at 250 °C and 300 °C, respectively.

It should be noted that the actual performance of a scaled-up reactor may not be exactly as predicted due to differences in mass and energy transfer between the bench-scale and pilot-scale systems. The effect of mass transfer limitations on the catalytic process was assessed via several methods [27]. First, bench-scale experiments were conducted using two separate reactors having internal diameters that differed by a factor of three. With each reactor, the production of CO₂ was measured under a range of W/F_v conditions, and plots of CO₂ produced versus W/F_v showed overlapping curves. An additional experiment was conducted, whereby the reactor diameter and mass of catalyst were both varied in a manner that yielded the same W/F_v condition, and the concentration of CO₂ produced by the two reactors differed by only ~6%. The experiments were purposely run at a flow rate that yielded less than 100% conversion to CO₂, and the results with the two different reactor configurations were 85% conversion vs. 89% conversion to CO₂. Finally, the pilot-scale reactor was designed in such a way that the linear velocity through the catalyst was within the range of linear velocities tested with the bench-scale experiments. Because of the similarity in linear velocities, it is assumed that the turbulent flow conditions inside the scaled-up reactor were similar to those present in the bench-scale experiments, suggesting that the bench-scale experiments were appropriately designed to predict the performance of the pilot-scale system.

2.4. Pilot-Scale Demonstration Studies

Based on the results presented above, a scaled-up catalytic reaction system was fabricated and installed for demonstration purposes at a large commercial brewing facility. A stream of CO₂ generated during the fermentation process containing gaseous ethanol (407 ppmv) and oxygen (~1800 ppmv) was passed through the catalytic reactor at an average flow rate of 100 L min⁻¹ (~25 lbm h⁻¹). The catalytic reactor was operated at a temperature of 300 °C. Given these treatment conditions, the mass of catalyst was chosen to provide 99% conversion at 300 °C under low O₂ conditions (i.e., 24 kg in Table 1) with a multiplier of 2 applied to account for any potential scale-up differences between the bench-scale and pilot-scale systems and to account for any potential effects related to operating the scaled-up reactions in CO₂, as opposed to nitrogen for the bench-scale studies. As such, 50 kg of catalyst was used in the scaled-up system. This system was initially operated intermittently for 18 days, during which time the operation was optimized. After optimization, the reactor was operated for 40 h cumulative run time over 6 days with CO₂ flowing, as described above.

The inlet and outlet daily concentrations of ethanol were measured via grab samples during the 6-day testing period. The inlet ethanol concentration ranged from 332 to 505 ppmv (average = 407 ppmv), and the percent conversion to CO₂ was above 99.9%, except for Day 3, which was at 99.6% conversion. The real-time measurements of inlet and outlet total hydrocarbons (THC) (mostly ethanol) and O₂ concentrations are shown in Figure 3 for the final day of operation (Day 6); THC and O₂ observations were similar for the other five days of the optimized test period. As can be seen, the THC concentrations were reduced from >400 ppmv to extremely low levels during most of the testing period. The catalytic process is sensitive to decreases in O₂ concentration. Slight increases in outlet THC levels around 4.75 and 7.25 h coincide with decreases in inlet O₂ levels. Slight increases in THC levels were also observed on other days, and manually increasing the inlet O₂ concentration using the auxiliary cylinder resulted in a subsequent rapid drop in outlet THC concentration to very low levels. Also shown in Figure 3, the catalytic degradation of hydrocarbons in the system is coincident with a drop in O₂ levels between the inlet and outlet. Stoichiometrically, a 400 ppmv decrease in ethanol concentration would require a simultaneous 1200 ppmv decrease in O₂ levels, which is roughly what was observed. Given the performance over the last several hours (hours 7 through 9) of operation, it was demonstrated that it is possible to optimize this system to simultaneously maintain very low levels of both THC (ethanol) and O₂. The measured O₂ and THC gas concentrations for day 6 at the inlet and outlet of the reactor are provided as Supplementary Materials (Table S3).

The inlet and outlet were also monitored for the presence of the specific organic compounds. In addition to ethanol, the following organic compounds were detected in the inlet grab samples (numbers in parentheses are average concentrations in ppmv during the six-day demonstration period): acetaldehyde (4.55), dimethyl sulfide (0.40), 1-propanol (0.18), ethyl acetate (10.1), 3-methyl-1-butanol (0.85), 2-methyl-1-butanol (0.33), iso-amyl acetate (3.19), ethyl hexanoate (0.59), myrcene (0.07), and ethyl octanoate (1.05). Methane was detected in the outlet grab sample on three of the six days of the demonstration period, at an average concentration of 4.4 ppmv. Benzene (0.014 ppmv), ethane (2.9 ppmv), methanol (0.090 ppmv), and toluene (0.029 ppmv) were detected in the outlet grab sample on the first day, and toluene was detected again on day two (0.010 ppmv). Finished product CO₂ quality impurity limits as set by the International Society of Beverage Technologists (ISBT) [28] are as follows for the compounds above: methane (total hydrocarbons), 50.0 ppmv; benzene and toluene (aromatic hydrocarbons), 0.020 ppmv; ethane (non-methane total hydrocarbons), 20.0 ppmv; and methanol, 10.0 ppmv. Qualitatively, brewery personnel observed the outlet gas to have a very clean aroma. Gas concentrations of reaction byproducts from the pilot-scale studies (experiments performed on Day 1 through Day 6) are provided as Supplementary Materials (Table S4).

To determine the extent of carbonaceous deposits on the surface of the catalyst after the pilot-scale study, a randomized representative subset (10% of the total) of the used catalyst was evaluated by careful visual examination, physical separation, and use of a laboratory balance. It was calculated that ~91% of the catalyst pieces exhibited no color change. Only about ~1% of the catalyst was black in color, indicating significant carbonaceous deposits. The remaining pieces (~8%) exhibited some color change. It should be noted that there is no direct evidence to show that these carbon deposits would result in a loss of catalyst activity.

2.5. Machine Learning and Artificial Intelligence for Catalysis Research

Machine learning (ML) is a form of artificial intelligence (AI) and continues to be an approach to accelerate catalyst research. ML learns from existing data to generate training models for predicting catalytic performance outside the training datasets. The trend of publications in ML-based catalysis research from the Web of Science database indicates that ML is being utilized increasingly [29]. Review articles have summarized studies in catalyst design and discovery and have discussed the significance of choosing suitable descriptors (also called features) [30]. Catalytic descriptors are representations of reaction conditions (e.g., catalysts and reactants) that are extracted from original data to describe target properties (e.g., conversion, yield, and selectivity) in a machine-readable form. Benavides-Hernández and Dumeignil [31] reported on the synergistic integration of ML and AI with high-throughput experimentation in heterogeneous catalysis, with a focus on catalyst characterization, data-driven exploitation, and data-driven discovery. Raccuglia and coworkers [32] demonstrated an approach that used ML algorithms trained on reaction data to predict reaction outcomes for a suite of materials, including metal oxides, metal–organic frameworks (MOFs), and perovskites. They used information from successful and unsuccessful (or “dark”) studies collected from archived laboratory notebooks and added physicochemical property descriptions to their raw notebook information using cheminformatics techniques. These data were used to train an ML model to predict reaction success, reporting that the ML model outperformed traditional human strategies and successfully predicted conditions for new templated inorganic product formation with a success rate of 89%. With approaches and techniques described in these studies, it is envisioned that the catalyst (Pt/TiO₂–ZrO₂), ethanol reactant conversion, and limited byproduct formation with a feed-stream of CO₂, as reported in the present study, can be used as training data in future ML/AI studies seeking to optimize catalyst formulation for a given desired outcome.

2.6. Future Work

While beyond the scope of this study, there are additional research questions that could be pursued in the future. First, the bench-scale study was conducted with N₂ as the balance gas (to allow for monitoring the CO₂ concentration to determine reaction “completeness”), while the pilot-scale study was conducted with CO₂ as the balance gas. Future work could focus on studying the effect of CO₂ vs. N₂ as the balance gas with regard to adsorption of relevant chemical species, reaction kinetics, and heat/mass transfer properties, to ultimately determine the effect on catalytic performance. In addition, temperature-programmed desorption (TPD) experiments with CO₂ (CO₂-TPD) could be conducted to identify site-specific interactions and to quantify the adsorption strength of CO₂ on the catalyst surface. Consider that Zheng and coworkers [33] conducted CO₂-TPD with a Fe-containing metal oxide catalyst (CeO₂) and (1) identified TDP peaks corresponding to weak, medium, and strong basic sites on the catalyst surface and (2) demonstrated that the adsorption of CO₂ by CeO₂ was enhanced by the introduction of Fe. Also, surface characterization studies could be conducted to identify intermediates observed on the surface of the used catalyst with regard to elemental makeup as well as solid form (graphitic vs. amorphous carbon, for example). A related study could attempt to determine if catalyst darkening leads to a decrease in catalytic performance, since this is only a hypothesis at this time. These future studies would help to improve catalytic predictions by improving the estimation of kinetic parameters and would provide practical insight into potential decreases in catalytic performance from the presence of adsorbed surface intermediates.

3. Materials and Methods

3.1. Catalyst

Sol–gel processing methods were employed to prepare TiO₂ and ZrO₂ suspensions via the acid hydrolysis of titanium isopropoxide (Gelest, Inc., Morrisville, PA, USA) and zirconium n-propoxide (Gelest, Inc., Morrisville, PA, USA), respectively [34,35,36]. Both suspensions contained approximately 2 wt% solids, with particle sizes of 2–4 nm diameter for ZrO₂ and 3–7 nm diameter for TiO₂, as measured by light scattering. These suspensions were mixed in a volume ratio of approximately 7:1 (TiO₂–ZrO₂). The resulting mixed-oxide suspension was diafiltered against water to raise its pH to 3. A portion of this suspension was then platinized by adding enough chloroplatinic acid to provide a 0.6 wt% platinum loading after reducing the adsorbed acid with sodium borohydride solution and performing an additional diafiltration step to remove the remaining boron salt. Specific properties of this TiO₂–ZrO₂ binary metal oxide material were described in a previous publication [35]. In that paper, the authors reported XRD spectra, BET surface area (~250 m² g⁻¹), percent porosity (~55%), electrophoretic mobility, and the effect of sintering temperature on phase transformations for a similarly prepared TiO₂–ZrO₂ solid.

The catalyst carriers (aluminum silicate) used for these tests were spherical with a 6 mm average diameter and a surface area of 30 m² g⁻¹ (Saint Gobain, Malvern, PA, USA). For the bench-scale tests, the carriers were dip-coated two times with the mixed-oxide suspensions and air-dried for at least one hour after each coating. These coated carriers were then dip-coated a third time with the platinized suspension and heated in air to a temperature of 350 °C for 4 h after coating. The mass of supported catalyst used in these experiments was either 1.9 g or 3.8 g. The same coating protocol was used to prepare the 50 kg of catalyst used for the pilot-scale tests, except that these catalysts were spray-coated at The Coating Place (Verona, WI, USA). No additional catalyst treatment was undertaken. From previous experience, three layers of catalyst have generally been determined to provide optimum conversion. It provides enough catalyst to facilitate the reactions without adding unnecessary thickness that would limit access of analytes to the interior of the coating. It is also thin enough to keep the catalyst layers firmly attached to the support material.

3.2. Bench-Scale Experiments

The bench-scale experiments employed the setup shown in Figure 4. A gas manifold was constructed to allow either zero-grade air (<1.0 ppm total hydrocarbons, Linde Gas, Murray Hill, NJ, USA) or an ethanol mixture (Linde Gas, Murray Hill, NJ, USA) to be passed through the heated catalyst. The ethanol mixture had an ethanol concentration of 360–503 ppmv, depending on the cylinder, with the balance nitrogen and O₂ (the ratio of nitrogen to O₂ was varied as specified below). Although the target application was the removal of ethanol from a feed stream of CO₂, reactions in the bench-scale portion of this study were conducted in nitrogen to enable measurement of CO₂ that was produced during the reactions. It is possible that the use of nitrogen instead of CO₂ as the bulk gas could somewhat affect the overall reaction kinetics, adsorption equilibria, and/or heat/mass transfer properties; however, the ability to monitor the reaction “completeness” by following the CO₂ concentration was determined to be an essential outcome of the bench-scale experiments, so nitrogen was chosen as the bulk gas.

Either the pure air or ethanol mixture was first routed through a preheater containing ¼ in. diameter type 302 stainless steel balls (McMaster Carr, Elmhurst, IL, USA), which were maintained at approximately 250 °C. After the preheater, the gas was routed into the catalytic reactor, maintained at a temperature of either 250 °C or 300 °C. The preheater and catalytic reactor both had a high-temperature thermocouple inserted, through sealed fittings, into the bed of stainless-steel balls or supported catalyst pieces. A gas sampling port was positioned immediately prior to the preheater and immediately after the catalytic reactor. Gas exiting the reactor was routed through a non-dispersive infrared gas analyzer (IRGA, described in Section 3.4), then into a flow meter (Model ADM-1000, Agilent Technologies, Santa Clara, CA, USA), and exhausted into a fume hood.

For a given experiment, the preheater and catalyst bed were allowed to reach the set temperature and equilibrated overnight with no gas flow. Prior to testing, pure air was passed through the system to precondition the catalyst—initial evaluation of the thermal pre-cleaning procedure is illustrated in Figure S3 in the Supplementary Materials section. Air flow continued until the reactor effluent was CO₂-free, as measured by the IRGA. The gaseous ethanol mixture was then flowed through the system at a predetermined rate (i.e., W/F_v ratio). Samples were periodically removed from the two sampling ports for analysis by gas chromatography. A time period of several hours was typically required for all concentrations to stabilize.

3.3. Gas Chromatography

A Hewlett-Packard 5890 Series II gas chromatograph with a flame ionization detector, GC-FID, was used for the bench-scale portion of this study. Helium was used as the carrier gas. For each analysis, 250 μL of the sample was injected using a gas-tight syringe. The injector was operated in the split mode with a split ratio of 25:1. The GC column was an HP-5 capillary column (25 m long × 0.2 mm i.d. × 0.33 μm film thickness, Agilent Technologies, Santa Clara, CA, USA) that was maintained at 30 °C for 4 min. The injection port and detector temperatures were maintained at 250 °C.

A Hewlett-Packard 5890 Series II gas chromatograph with a mass selective detector (GC-MSD) was used to identify gas-phase intermediates. Helium was used as the carrier gas, and the injector was operated in the split mode. A 250 μL injection of the sample was performed using a gas-tight syringe. The GC column was an SPB-5 capillary column (30 m long x 0.25 mm i.d. × 0.25 μm film thickness, Supelco, Bellefonte, PA, USA) that was maintained at 30 °C for 4 min. The injection port temperature was 120 °C, and the MSD transfer line was maintained at 280 °C. The MSD was operated in the selected ion monitoring (SIM) mode to increase the sensitivity.

3.4. Non-Dispersive Infrared Gas Analyzer

In the bench-scale experiments, a Qubit (Kingston, ON, Canada) Model S151 IRGA was used to measure the concentration of CO₂ exiting the reactor. For these experiments, the IRGA was positioned in-line between the reactor outlet sample port and the flow meter. A small column packed with a water-adsorbing solid (magnesium perchlorate) was placed before the IRGA to eliminate the effect of water vapor on the measurement of CO₂.

3.5. Scale-Up Predictions

To allow for appropriate pilot-scale sizing, the bench-scale catalytic experiments had to be conducted at a variety of volumetric flow rates for a given set of test parameters (i.e., temperature and O₂ concentration). The data generated in this study were then fit according to a pseudo-first-order model, with a reaction rate expression of the form:

$$r_A={\displaystyle \frac{V}{W}}{\displaystyle \frac{\mathrm{d}{C}_A}{\mathrm{d}t}}=-k_A{C}_A$$

(2)

where r_A is the analyte reaction rate (in units of mol g⁻¹ s⁻¹), V is the void volume of the reactor (in units of L), W is the catalyst mass (in units of g), C_A is the analyte concentration (in units of mol L⁻¹), t is the time (in units of s), and k_A is the analyte reaction rate constant (in units of L g⁻¹ s⁻¹). Other reaction orders were investigated, including 1/2, 3/4, 3/2, and second, and the first-order model provided the best overall fit. The excellent fit of the first-order model to the experimental data is shown in the Results and Discussion (Section 2.1). Note that the authors have utilized the aforementioned kinetic reaction order models, as well as a Langmuir–Hinshelwood–Hougen–Watson model, in previous studies [36].

Equation (2) can be integrated between the limits of C_A0 and C_A to yield the following linearized form of a first-order model:

$$\ln C_A=\ln C_{A0}-k_A{\displaystyle \frac{W}{F_v}}$$

(3)

where C_A0 is the initial analyte concentration (in units of mol L⁻¹), and the ratio V/t has been substituted by F_v (the volumetric flow rate, in units of L s⁻¹). The reaction rate constant was determined by performing a least-squares regression analysis of ln C_A versus W/F_v, where k_A is equal to the negative of the slope.

Based on the difference between the amount of ethanol degraded and the amount of CO₂ produced, it was apparent that the catalytic reactions in this study produced a significant amount of unmeasured, non-volatile intermediate byproducts under certain flow rates. It is likely that most of these intermediates remained on the surface of the catalyst (primarily as carbon deposits) because they were not detected in the gas-phase effluent (unless otherwise noted below), and the catalyst was visibly darkened after some of these experiments. This carbonaceous material was observed to be insoluble in water, ethanol, or hexane. Upon heating the darkened catalyst in the presence of flowing pure air, a significant amount of CO₂ was measured with the inline IRGA, further suggesting that carbonaceous surface intermediates were present. The general term “intermediates” will be used in this manuscript to refer to any gas-phase intermediates plus this carbonaceous material, since the ultimate goal is complete mineralization to CO₂, and the reaction sequence will be modeled as follows:

Ethanol → Intermediates → Carbon Dioxide

The reaction rate of each of these species was modeled as follows:

Ethanol (E)

$$r_E={\displaystyle \frac{V}{W}}{\displaystyle \frac{\mathrm{d}{C}_E}{\mathrm{d}t}}=-k_E{C}_E$$

(4)

Intermediates (I)

$$r_I={\displaystyle \frac{V}{W}}{\displaystyle \frac{\mathrm{d}{C}_I}{\mathrm{d}t}}=k_E{C}_E-k_I{C}_I$$

(5)

Carbon Dioxide (C)

$$r_C={\displaystyle \frac{V}{W}}{\displaystyle \frac{\mathrm{d}{C}_C}{\mathrm{d}t}}=k_I{C}_I$$

(6)

For each of the species, the predicted concentration at a given W/F_v ratio was subsequently modeled as follows (where C_E0 is the initial concentration of ethanol) to evaluate the appropriateness of the calculated reaction rate constants, as well as the linear model:

Ethanol (E)

$$C_E=C_{E0}{e}^{-k_{E}W/F_v}$$

(7)

Intermediates (I), see Hill [27]

$$C_I={\displaystyle \frac{k_E{C}_{E0}}{k_I-k_E}}(e^{-k_{E}W/F_v}-e^{-k_{I}W/F_v})$$

(8)

Carbon Dioxide (C), see Bailey et al. [37]

$$C_C=C_{E0}\left[1-{\displaystyle \frac{k_I{e}^{-k_{E}W/F_v}}{k_I-k_E}}+{\displaystyle \frac{k_E{e}^{-k_{I}W/F_v}}{k_I-k_E}}\right]$$

(9)

3.6. Pilot-Scale Installation and Operation

Based on the analysis described above from the bench-scale experiments, a pilot-scale catalytic CO₂ treatment system was designed, fabricated, and installed at a commercial brewing company to perform technology demonstration studies. A process flow diagram of the pilot-scale CO₂ treatment system is given in Figure 5. The salient hardware comprising the pilot-scale CO₂ treatment system included (a) a single-stage, 2-cylinder gas compressor and horizontally-positioned tank-style gas reservoir; (b) an auxiliary O₂ gas cylinder (Airgas, Madison, WI, USA); (c) an insulated pre-heater; (d) a large, insulated (jacketed) vertical reactor containing 50 kg of supported catalyst; and (e) a water-cooled heat exchanger. Pressure relief valves were affixed atop the gas reservoir and the reactor to protect the equipment from excessive pressure. The flow rates of CO₂ and the auxiliary O₂ feed were controlled and monitored via flow meters connected to a 4-channel display. Two manually controlled back-pressure valves maintained desired pressures throughout the treatment system. Pressures at select locations in the treatment system were monitored via transmitters coupled with a power supply and display, while temperature transmitters consisted of Type J thermocouples. Pressure indicators (gauges) and flow indicators (rotameters) provided a visible means of monitoring system operation at select locations. A data acquisition (DAQ) system (CompactDAQ hardware, LabView software, National Instruments Corp., Austin, TX, USA) was used for electronic monitoring and control of pressures, temperatures, and flow rates in the treatment system.

Samples of CO₂ gas for analyses were acquired at the reactor inlet and outlet (Figure 5) of the reactor. Each site utilized two gas sampling approaches: (a) grab samples via Summa canisters for subsequent laboratory analysis (Airborne Labs International, Somerset, NJ, USA) and (b) transfer of gas to an in-line, real-time total hydrocarbon analyzer (Model: 51C-HT, ThermoFisher Scientific Corp., Franklin, MA, USA) and a fuel-cell-based O₂ analyzer (Model: GPR-1600 MS Series PPB Oxygen Analyzer, Advanced Instruments Inc., Norwood, MA, USA). Sample transfer tubing between the inlet and outlet sites and their respective THC and O₂ real-time instruments was electrically heated and insulated to prevent condensation in the lines. Calibrations of the THC and O₂ monitoring instruments were performed periodically throughout the study with cylinder gases containing certified concentrations, per manufacturer specifications.

A stream of CO₂ generated during the fermentation process and containing gaseous ethanol at an average concentration of 407 ppmv was passed through the catalytic reactor at an average flow rate of 100 L min⁻¹ (~25 lbm h⁻¹). The average O₂ concentration in the feed stream was 1800 ppmv, and the catalytic reactor was operated at a temperature of 300 °C. The pilot-scale system was operated for a period of 18 days, allowing for the system and processes to be conditioned, studied, and optimized, followed by another period of 40 h (over the course of six days), during which data were collected.

4. Conclusions

In this study, the gas-phase catalytic degradation of ethanol was first examined on a bench scale, and the effects of temperature and oxygen concentration were considered. Ethanol was completely converted to CO₂ under low flow rate conditions, and the reactions proceeded through volatile and non-volatile reaction intermediates. Models were used to design a pilot-scale catalytic treatment system. The pilot-scale unit was fabricated and installed at a large commercial brewing company to treat a slip stream of CO₂ feed containing ethanol vapor and O₂. The pilot-scale system reduced the feed stream ethanol concentrations by 99.9% while concomitantly reducing the O₂ concentrations over the course of a six-day demonstration period without generating unacceptable levels of byproducts.

Based on the bench-top and pilot-scale observations reported herein, the hypothesis stated in the introduction is supported. Using the catalytic treatment system described in this manuscript, a feed stream of highly purified CO₂ could be collected by a brewery for internal reuse or for external sale. If the pilot-scale treatment system were operated on an annual basis with typical time periods associated with the brewing process, the mass of CO₂ exhausted to the ambient could be reduced by an estimated 2.6 tons of CO₂ per year. Expanding the pilot-scale to a full-scale system would provide an even larger environmental benefit. Companies and their products have a carbon footprint. The possibility of accumulating carbon credits due to reduced CO₂ emissions would be a supplemental benefit to breweries employing this technology. The catalytic treatment system could be rescaled (up or down) for other applications (e.g., non-beverage ethanol fermentation) after conducting bench-scale testing and modeling akin to those described herein.