Modeling the Impact of Solid Surfaces in Thermal Degradation Processes

Glycerol (GC, propane-1,2,3-triol), propylene glycol (PG, propane-1,2-diol), and triacetin (TA, propane-1,2,3-triyl triacetate) (Figure 1) are commonly used as solvents, food additives, and humectants in the food and tobacco industries. Their use often involves high-temperature conditions, in which thermal decomposition reactions occur, exposing the human body to potentially harmful substances formed during pyrolytic processes. For this reason, it is of great importance not only to know their pyrolysis but also to study the underlying reaction mechanisms. A detailed picture of the energetics and kinetics of the pyrolytic process is the key to making reliable predictions for the products formed under certain external conditions.

Figure 1. Chemical structures of glycerol (GC), propylene glycol (PG), and triacetin (TA).

Recently we investigated the pyrolysis of GC, PG, and TA in the gas phase. For GC, we provided a novel overall decomposition scheme whose characteristic is the formation of glycidol as the first intermediate representing the overall rate-limiting step.1 The investigation of PG dehydration led to the discovery of propylene oxide as a similar intermediate during the rate-limiting step.2 Regarding TA, we presented a scheme describing its complex thermal decomposition pattern for the first time, in which the rate-limiting step is predicted to consist of the initial elimination of acetic acid.2 For both PG and TA, we characterized major decomposition products and the corresponding energetics, complementing theoretical results with qualitative data from experiments. Unfortunately, for all the aforementioned substances, a comparison with published experimental data, if available, is a complex task. Experimental studies are always performed under specific catalytic conditions that are totally different from a clean pyrolytic gas-phase scenario.

Gas-phase simulations of pyrolytic steps offer ideal conditions to study a broad range of possible chemical degradation processes in detail. However, it is also clear that carbohydrate and fat pyrolysis chemistry occurs widely wherever biomass is combusted or food is cooked. These rather complex situations can hardly be simplified by investigating such chemical compounds under pure gas-phase conditions. In fact, the presence of solid particles and the interactions of organic compounds with a variety of solid surfaces may strongly affect the energetics of the degradation processes, leading to pyrolytic patterns that are totally different to what is known for the gas phase.

Several experimental studies related to surface catalyzed degradation reactions of GC have been reported,3–12 describing optimal conditions to convert GC either into acrolein or acetol. One reason for these extensive investigations is that GC became available widely as a byproduct of the transesterification of fats into biodiesel. Regarding the reactivity of GC in the presence of catalytically active surfaces, only few computational studies exist in the literature: density functional theory simulations on the reactivity of GC over zeolites13 and transition metals.14 More recently, Vlachos and coworkers15 reported a QM/MM investigation on the acid catalyzed dehydration of polyols in liquid water. However, to the best of our knowledge, neither experimental data nor computational results for the pyrolysis of PG and TA over solid surfaces/materials are available in the literature.

Nonetheless, the information available from publications in this field is relevant only to conditions extremely different from those found in everyday life. The latter involve the exposure of GC, PG, and TA to carbon and metal-oxide particles, as in cigarettes,16 or when pots and pans are used in high-temperature food-preparation processes. For this reason, our study focuses on the rate-limiting steps of the gas-phase decomposition of GC, PG, and TA predicted in our recent works1, 2 and simulates how barrier heights change when these reactions occur at the surfaces of different solids, characterizing the key features that are relevant for thermal degradation processes. Our candidates for the solid materials are amorphous carbon, amorphous silica, crystalline zirconia, and crystalline alumina. Computational details and methods are available in the Supporting Information.

The pseudo-amorphous carbon surface contains reactive sites (low-valent carbon atoms). They interact with oxygen atoms of the GC hydroxyl groups. With a barrier of 19 kcal mol⁻¹, primary hydroxyl groups are split, forming an energetically more stable surface oxide intermediate (chemisorption). This intermediate is connected to a barrier height ΔE^≠ of 25 kcal mol⁻¹ for the formation of glycidol, see Figures 2 and 3. The corresponding transition structure formed after CO bond splitting is similar to a primary carbocation, and its relative stabilization when compared with the situation for the unimolecular reaction step in the gas phase1 (ΔE^≠=59 kcal mol⁻¹) can be attributed to the energy released by incorporating the leaving oxygen atom into the amorphous carbon framework. This is the main effect, which even leads to a negative apparent barrier height, that is, the energy for the adsorption of GC from the gas phase at the reactive site of the amorphous carbon surface is larger than the barrier height for the conversion of adsorbed GC to glycidol.

Figure 2. DFT (PBE+D) total energy reaction profiles for the decomposition of glycerol in the gas phase and at the surface of various materials (numbers in kcal mol⁻¹).

Figure 3. Decomposition of glycerol (column 1) to glycidol and water (column 3) over amorphous carbon (row 1), pseudo-amorphous silica (row 2), crystalline zirconia (row 3), and crystalline alumina (row 4) via the transition structures shown in column 2 (atom colors: red—O, white—H, blue—C, yellow—Si, green—Zr, purple—Al).

GC forms hydrogen bonds with surface oxygen atoms of silica and zirconia (Figure 3). The corresponding PBE+D physisorption energies are −27 kcal mol⁻¹ in the case of amorphous silica and −20 kcal mol⁻¹ for crystalline zirconia. Here, desorption competes with the activation process for the formation of glycidol. The intrinsic total energy PBE+D barrier heights are 59 kcal mol⁻¹ (silica) and 62 kcal mol⁻¹ (zirconia), that is, two to three times the energy required for the desorption of GC. Furthermore, these intrinsic barrier heights are very close to that for the same reaction step in the gas phase (59 kcal mol⁻¹).

The clean alumina (0001) surface, however, turns out to be more reactive than the other oxide material surfaces. This is due to the Lewis acidic character of the surface aluminum atoms. As already shown in previous studies related to alumina (0001) water adsorption,17, 18 a barrier of only a few kcal mol⁻¹ is associated with the proton transfer from physisorbed water molecules to surface oxygen atoms and the simultaneous creation of surface AlOH groups. GC hydroxyl groups exhibit a slightly higher Brønsted acidity than water and, as a consequence, lead to a direct (i.e., barrierless) chemisorption of GC via those hydroxyl groups that do not form an internal GC hydrogen bond. For GC adsorbed via a primary, its secondary, or both a primary and the secondary hydroxyl group, we obtain adsorption energies of −36, −40, and −71 kcal mol⁻¹, respectively. In these configurations, the secondary hydroxyl group always forms an internal hydrogen bond with a free primary hydroxyl group, resembling the structural entry point for the formation of glycidol. Corresponding barrier heights are 45 kcal mol⁻¹ if starting from the adsorption via the secondary hydroxyl group only and 50 kcal mol⁻¹ when GC is adsorbed via both the secondary and a primary hydroxyl group. In the latter case, we also arrive at a negative apparent barrier height, a situation similar to the case of amorphous carbon, see above.

Similar to GC, PG forms hydrogen bonds to surface oxygen atoms of silica and zirconia. The intrinsic barrier heights for the formation of propylene oxide are virtually identical to the corresponding gas-phase value of 59 kcal mol⁻¹, see Figure 4. Again, this is two to three times the physisorption energy, rendering the desorption of PG more likely than its decomposition at the surface.

Figure 4. DFT (PBE+D) total energy reaction profiles for the decomposition of propylene glycol in the gas phase and at the surface of various materials (numbers in kcal mol⁻¹).

A much stronger attraction between PG and the material surface is predicted in the case of alumina (−44 kcal mol⁻¹). Adopting a configuration suitable as precursor for propylene oxide formation, the PG secondary hydroxyl group in the adsorption structure both acts as electron donor for an aluminum surface atom and forms an internal hydrogen bond with the primary hydroxyl group. A structure very similar in nature, associated with a barrier height of 50 kcal mol⁻¹ for the epoxide formation, was already found for GC chemisorbed at the alumina surface (see previous subsection). This is consistent with the value of 48 kcal mol⁻¹ found for the intrinsic barrier height of PG to propylene oxide at the alumina surface.

As for GC, the smallest physisorption energy of PG at the material surface is obtained for amorphous carbon (−8 kcal mol⁻¹). However, overcoming a modest barrier of 15 kcal mol⁻¹, a much more stable surface oxide can be formed (chemisorption), such as in the case of GC. Accordingly, this intermediate is the precursor for the formation of the epoxide structure connected to a barrier height of 24 kcal mol⁻¹. There is, however, a lower lying transition path, with a barrier height of only 19 kcal mol⁻¹, leading to acetone which is also energetically more stable than propylene oxide. Here, the proton of the secondary hydroxyl group is transferred to the surface framework, and a hydride transfer from the secondary to the primary carbon atom, which previously splits off from the surface oxygen bridge, completes the formation of acetone. Note that this reaction step represents a shortcut compared with the decomposition in the gas phase, where the formation of acetone involves the formation of propylene oxide as its precursor2 together with an intrinsic activation energy about four times higher than the one observed here starting from the surface intermediate.

TA contains ester groups which are less reactive than the hydroxyl groups of GC or PG. This reduces its interaction with the amorphous carbon surface to a level as weak as dispersion. Moreover, the carbon surface does not even exhibit partial ionic character, such as that of oxide materials, and therefore, it can be expected that the decomposition of TA is hardly affected by the presence of amorphous carbon. However, attractive forces exist between carbon atoms of the ester groups and the surface oxygen atoms, as well as between oxygen atoms of the ester groups and the Lewis acidic aluminum atoms of the alumina surface. Including van der Waals interactions, the TA physisorption energies for silica, zirconia, and alumina sum up to −10, −19, and −28 kcal mol⁻¹, respectively, see Figure 5. The corresponding intrinsic barrier heights for the decomposition of TA into propene-1,3-diol diacetate and acetic acid are slightly below the value of 43 kcal mol⁻¹ obtained for the gas phase.2 This reduction is due to a transition structure stabilization effect resulting from an interaction with the polar oxide surface. This effect is less pronounced for alumina than for silica and zirconia because here the structural flexibility of the TA molecule at the surface is smaller owing to stronger binding effects.

Figure 5. DFT (PBE+D) total energy reaction profiles for the decomposition of triacetin in the gas phase and at the surface of various materials (numbers in kcal mol⁻¹).

In summary, we performed first-principles simulations for the initial steps of the thermal decomposition of GC, PG, and TA over the surfaces of pseudo-amorphous carbon and silica, crystalline zirconia, and crystalline alumina. Our results show that surfaces that do not contain reactive sites hardly influence the decomposition behavior. We have seen, for example, that in the case of silica and zirconia, the intrinsic barrier heights virtually do not change with respect to the values obtained for the gas phase. Furthermore, the adsorption energies are much smaller than the intrinsic activation energies, which renders desorption more likely than a decomposition event at the material surface. The picture changes when the adsorption is controlled by active surface sites. In the case of the clean alumina surface, we observed a strong interaction between Lewis acid aluminum atoms and the hydroxyl groups of GC and PG. On the one hand, this leads to a strong binding between the alcohol molecule and the surface (chemisorption); on the other hand, it can increase the acidity of the hydroxyl groups such that the barrier heights of the epoxide formation decrease. Overall, this results in a significant speedup of the decomposition of GC and PG with respect to the gas phase. Of all the oxide material surfaces investigated in this work, the alumina surface exhibits the largest TA adsorption energy. Although the intrinsic barrier for the decomposition into propene-1,3-diol diacetate and acetic acid remains virtually constant at all oxide surfaces, in the case of alumina the energy required for the desorption of TA is high enough for the decomposition step to become competitive. As a consequence, an acceleration effect for the decomposition of TA can be expected at the clean alumina surface. The effects induced by the surface of amorphous carbon differ from those of the oxide materials: Low valent carbon atoms induce the chemisorption of GC and PG, forming surface alkoxides. They not only exhibit a significantly lower barrier for the formation of the epoxides, but also open the door to decomposition mechanisms not predicted for the gas phase.