A research into the thermodynamics of methanol to hydrocarbon (MTH): conflictions between simulated product distribution and experimental results

Initial work–thermodynamic calculations on selected sub reactions

The calculations on the thermodynamics of individual sub reaction are included as the basis work of, and part of the essential studies referring to the later analysis. For such calculations, one disadvantage that may concern in the prediction of real experimental results is that assumptions have been made that each sub reaction proceeds without the influence of others, although they may be in a great certain to proceed in a parallel way in a real MTH system. At this point, previous studies recalled computer aided simulations which have integrated all the possible sub reactions and, therefore, might give a more comprehensive result [8, 9]. In this work, we have achieved a more comprehensive computer-aided (CA) simulation from which the results have integrated all the possible influences between different sub reactions, and, therefore, might give a more indicative result for real industrial applications. Advantages of such a method base on the ability to solve chemical system (i.e. MTH system) proceeding in “black-box” of uncertainty. Notably, in reality it is impossible to have the MTH reaction finish completely into the form of cokes; on the contrary, similar to the lab experiments, any MTH reaction in an industry-scale reactor or pilot instrument would be a non-stopping, products-continuously being-removed process, and particularly when the product selectivity obtained by the employed porous zeolites (e.g. H-ZSM-5) works to affect the yields, the calculations on single sub reaction do denote important implications for the product compositions achieved in various conditions.¹

Computer-aided simulations for three reactions of different conditions

Equilibrium compositions of product mixture in the temperature range of 200–1000 °C are simulated for reactions under different reaction conditions for comparison with the lab test results. The simulations were employed by ASPEN Plus process simulation with the RGibbs model. The computer-aided simulation possesses many advantages when applied to a complex system in which multiple reactions processing in a parallel way, and the minimum reaction Gibbs free energy calculation has overcome the difficulty caused by the un-certained reaction routes and the interferences between different sub reactions [9, 10].

One unique important difference between the results of computer-aided (CA) simulation and the real lab-experiments is the extent of reaction process that might be taken into the calculation. Beyond all doubts, the CA simulation has been based on the assumption that the reaction has finally reached to the ‘extreme’ equilibrium; for instance, the methanol conversion happens in a close system for a termless period, thus, the layout of the results somehow reflect an ‘ideal’ product distribution which may never occur in reality. On the contrary, the lab tested results and even the real industrial plant outputs are obtained from a ‘non-stoping’, and continuos flow system. It is nearly impossible to have the whole system to reach a ‘final’ status (all cokes in this case), although some coking positions might have reached a close status. In our study, coke formation was not involved for results close to a ‘processing’ product output; however, it was found even CO, CO₂ were predominant in the simulated product composition, and this is similar to the case that coke is involved (Fig. S1). Obviously, the selectivity of our catalysts and the limited dwell time of primary products in the catalyst system gives a very small chance for CO and CO₂ to occur in the final product yield. The process simulation results without CO and CO₂ show closer to the real lab observations (Fig. S2), but the product distribution is limited to aromatics mainly. These problems in thermodynamic analysis, caused by extremely thermostable, but un-desirable products occurred in previous studies, and a rational solution normally prefer to ignore them in the calculations, for results better referenced to real lab observations [9, 10].

Our optimized simulation targets on the yield of major gas olefinic products (MTO) and aromatics (MTA) from MTH reactions in a ZSM-5 catalyst system. Former study obtained results closer to real experimental data when employed analogous methods [9]. However, our assumption is based on the ‘hydrocarbon-pool’ cycle, which has been considered as the origin of series of MTH reactions and gives C ₂ –C ₄ olefins and methyl-aromatics as the primary products [4]. Further consideration is given to the fact that latter stage hydrogen transfer may convert these primary products into either saturated hydrocarbons (alkanes) or graphite species (deeply dehydrogenated cokes), and these intra-conversions indeed play an important role in deciding the following catalytic process, and the final product yields [3, 4, 11, 12]. This simulation also takes the advantage that all of these ‘hydrocarbon-pool’ primary products (C₂–C₄ olefins and C₆–C₉ aromatics) take a large portion in the product yield of short time (a few minutes to several hours) lab reactions and industrial applications [4, 5].

In the present research, we have simulated 3 individual MTH reactions of different conditions in a wide temperature range of 200–1000 °C, and matched them with data from corresponding lab experiments (at single temperature points). Figure 1 shows the overall temperature dependent product distributions of a selected reaction (6 bar pressure, methanol/N₂ = 1:2 in mole), and results of other two reactions are included in the supplementary (for the three reactions, simulations with gas alkanes, and CO as well as CO₂ are also included, see Figs. S3–S8).²

Open image in new window

The simulated product distributions of different reactions are quite similar, showing a general trend of increasing selectivity to aromatics as the reaction temperature goes up. The single benzene ring aromatics always take the largest portion in the product, and methyl-benzenes (Toluene, xylenes and tri-methyl benzenes) reached a maximum selectivity in the temperature section of 300–400 °C, whereas benzene become predominant in the product distribution at temperatures of 400 °C and above.

The Fig. 2a–c show part of the products distributions without data of benzene and toluene, in the most widely applied methanol conversion temperatures (200–550 °C). It was found methanol concentration in feedstock (reflected by the ratio to N₂) only has limited impacts on the simulation data. The two reactions under atmosphere pressure exhibit quite close results with tiny difference (Fig. 2b, c). Notably, higher pressure caused observable shift of maximum product selectivity from the higher temperature zone to lower temperatures. An apparent evidence is the maximum M-xylene (MX) selectivity under 6 bar pressure, occurs at below 250 °C (Fig. 2a), whereas the corresponding maximum (as reflected by the peak centre in the figures) under atmosphere pressure occurs at about 275 °C (Fig. 2b, c). Our simulation well supports the real experimental observations that higher pressure and a lower temperature may benefit the yield of aromatics [2, 11, 17, 18].

Open image in new window

Inhibition of light olefin products has been reflected in the results (Fig. 3a–c) obtained in each reaction, showing formation of olefins is not thermodynamically favoured under the applied conditions. In the temperature range of 300–500 °C, which was indicated to donate large portions of C₂–C₄ olefins in the previous researches, [3, 5, 12, 19], ethylene selectivity is higher than other two olefin types, and shows a gradually increasing trend. It seems C₄H₈ (1-butene) is greatly supressed in each case, as reflected by the observed data. Propylene selectivity experienced a peak at about 300 °C, which is slightly higher under lower pressure conditions. Notably, all C₂–C₄ olefins are further suppressed by 6 bar pressure (Fig. 3 a).

Open image in new window

Conflictions between the simulated data and real experimental results

Simulation in this study was originally designed for the prediction of experiments and later explanations. Interestingly, the CA simulated results are quite different from the real experimental data. In each case, aromatics are always dominant in the simulated product distributions, especially benzene, reaches up to ~70 mol% in the case of 6 bar pressure, and 450 °C reaction temperature with a methanol to N₂ molar ratio of 1:2. Simulation shows higher pressure and smaller methanol to N₂ ratio both contribute to higher benzene selectivity, and thus possess hindrance on other aromatic products and olefins. Toluene is also a favourable product type by the thermodynamic and takes a large portion in the simulated data. It seems that methyl groups on the aromatic ring structure have increased the reaction difficulty thermodynamically, as product selectivity decreases apparently from benzene to tri-methyl benzenes; this exists in the simulated data of all the three reactions.

Another important information from the above simulated data is the extremely poor olefin selectivity (less than 1 mol%) in strong comparison to the aromatics. However, for the selected three different MTH reactions, gas products always take a large portion in the real product distribution. In the reaction under 6 bar, and with a lower methanol conversion duty (WHSV = 1.5 h⁻¹), aromatics yield has been promoted, but gas olefins and alkanes are still the major products. Observed yields of ethylene and propylene lie on a relatively low level, but still contribute to a total 3.77 mol% in the product distribution. Reactions under atmosphere pressure donate much higher level of olefins, although other conditions are different in the selected two reactions. Notably, higher methanol to carrier gas (N₂) ratio may also contribute to the improved olefins yield, as shown in the simulated results of two atmosphere pressure reactions, with methanol to N₂ ratio of 50:1 and 3:1, respectively. It is supposed that pressure is still the main reason for the totally different aromatic and olefin product distributions in each case, which is also supported by series of previous studies [1, 3, 4, 5, 19].

Simulations and calculations vs. limited dwelling time

Questions come from the huge difference in selectivity to gas olefins and aromatics between the simulated data and results from real MTH experiments. Obviously, the thermodynamic data are based on a perfect equilibrium status which may require an extremely long term reaction. Nevertheless, such a termless reaction, processing automatically, never occurs in both lab experiments and industrial applications, and it is nearly impossible to let MTH reaction be completely finished in a perfect status. The real reaction normally employs catalysts to promote certain steps, or some sub-directions of the whole reaction. Despite that the final equilibrium will never be disrupted by the catalyst, the reaction routes can be changed, and the product distribution in output are, therefore, totally different from the thermodynamic favoured answer, especially when the reaction time is limited. Here the limited reaction time means a short dwelling time of the processing products in the system, which cannot be converted to the final form before being transported to the outside environment. This is the reason why benzene is not allowed to be dominant in the real reaction products, although it is greatly thermodynamically favoured by the equilibrium. It also helps in understanding of the much higher olefin yields, which may not be chosen by the thermodynamic conditions, but play an important role as intermediate products. Conflictions lying between the simulation data and experimental records, therefore, have highlighted the importance of reaction routes which decide the product yield from the system in a parallel way with the thermodynamics in the case that the reaction cannot be ‘completely finished’ thermodynamically, and actually, they indeed have some more influences on product selectivity when the reaction time is limited.³

In the hydrocarbon pool mechanism, if alkylation steps on benzene ring are considered as the ‘starting point’ of series reactions, the following steps would be the corresponding ‘de-alkylation’ during which the alkyl groups on the benzene ring structure are eliminated, released to the reaction system in terms of the primary light olefins [12]. These light olefins may be converted into alkanes, or form more aromatics via the aromatization steps in the later stages. It is supposed that in practical laboratory tests and industrial manufactures, olefin products are transported quickly to the outside environment, and present in the yields, thus have limited chance to be converted into aromatics. On the other hand, some portions of the primary olefins are converted into aromatics, with higher pressure putting a positive effect on this process. The thermodynamic simulation is closer to the case that the system is isolated with endless reaction time towards a perfect equilibrium. In the case that coke formation is selectively ignored (the formed aromatics are not further converted into the graphitic cokes), the alkylation step is first forced to be interrupted, as no methyl group from methanol is supplied. However, de-alkylation may not stop immediately, and more light olefins are released from the alkyl benzenes, then, converted into more aromatics in the later stage. As no methyl group is supplied from outside of the system, the aromatization of light hydrocarbons would stop at the benzene ring structure. As a result, benzene finally becomes predominant in the system and methyl aromatics have less methyl groups have more chances to exist in the final product. When the hydrocarbon pool mechanism is introduced for explanations on reaction route, the thermodynamic simulation data and real experimental results could be better matched.

On the other hand, the limited dwelling time further reduces the chance of the intra-conversions (H₂ transfers) between the primary products which have been considered to possess unique impacts on the final product distributions [3, 4]. Accordingly, the selectivity to a single product may more depend on a ‘straight reaction route’. Our initial calculations on Gibbs free energy change indicate the thermodynamic barrier for different individual product formations directly from methanol (Fig. 4). These calculations have simplified the reaction by supposing only one target product is formed with only methanol and ignored product interactions. All the \(\Delta_{r} G_{m}^{\theta }\) of different reactions are less than 0 in the temperature range from 200 to 600 °C, which indicate that all the reactions are spontaneously in these temperature. The process simulation has considered all the reaction types including the interactions between different products (e.g. the transformation between different aromatics in Fig. 5).

Open image in new window

Open image in new window

The shape selectivity of different porous catalysts should also be considered. In a real MTH reaction over zeolite catalyst, e.g. H-ZSM-5, the product yield possesses large portion of gas olefins and alkanes, although they are greatly suppressed in the thermodynamic simulations. The channel diameter of ZSM-5 zeolite is close to the molecule size of single benzene ring aromatics, [4], thus limits the formation of larger aromatics. This also gives more opportunities for light molecules such as C₂–C₄ olefins to escape from the system in priority, and reduces the chance to be converted into aromatics. Gibbs free energy calculations on individual sub reactions show that methanol to olefins is thermodynamically favoured when each of the 25 sub reactions is isolated, in which case propylene and butane are easier to be formed than ethylene. When each sub reaction is isolated, the transformation of olefins to corresponding alkanes is also thermodynamically promoted, but gradually becomes inhibited at higher temperatures. Assumptions are made that the shape selectivity of zeolite allows selected reactions (e.g. methanol to olefins) to be more isolated. The related results are shown in the Fig. 6.

Open image in new window

Reaction pressure vs. product selectivity

In the above discussions, all reactions employed a fixed pressure. By comparing simulations under 6 bar and atmosphere pressure (Figs. 2, 3) we noted higher pressure may benefit to the formation of aromatics while put suppressions on the olefins generation. When the reaction temperature is fixed, the influence of reaction pressure becomes more apparent, in both simulation data (Table 2) and experimental results (Table 3). Therefore, we have simulated the MTH reaction equilibrium selectivities of important olefins (ethylene, propylene and butylene) and aromatics (PX, MX, and OX; 123-trimethyl-benzene, 124-trimethyl-benzene and 135-trimethyl-benzene: benzene and toluene) products under the pressures of 0–10 bar, taking 2 bar as the gap between each simulation (Figs. 7, 8, 9, 10). Methanol is the only reactant and assumptions are made that the selected olefins and aromatics are the only products of system. In each figure, the curves show the selectivity change of a particular product as the temperature changes in 200–1000 °C under different pressures.

Open image in new window

Open image in new window

Open image in new window

Open image in new window

The results indicate higher pressure increases selectivity to the C₂–C₄ olefins, but the values are still quite small. Under each pressure, ethylene selectivity goes up with the temperature while butylene drops in a dramatic trend and nearly disappears at temperatures of 600 °C and higher. On the other hand, propylene selectivity experienced a maximum at 300−500 °C, and higher pressure obviously shifts this maximum from lower temperatures to higher. For xylenes, their selectivity also experienced a maximum as shown by the peak in the temperature range of 200–450 °C. Interestingly, higher pressure does not promote higher selectivity to them, but just shift the maximum selectivity to higher temperatures. This is also the case for toluene. For tri-methyl benzenes, they do have higher selectivity when pressure goes up, but the general trend is, the selectivity drops rapidly as the temperature increases under a particular pressure. Similar to the early simulations, benzene always takes the largest portion in the product, its selectivity increases dramatically with the temperature, higher pressure just makes the maximum present earlier.

The above simulation results are closer to the real MTH reactions proceeding in time-on-stream, although the olefin selectivity is still tiny. Simulations considered alkanes, as well as CO and CO₂ are also included in the supplementary data. Again, the results show either only aromatics or CO and CO₂ are predominant in the products.