Carbon number distribution, olefin

The solvent is 28 CC-olefins recycled from the fractionation section. Effluent from the reactors includes product a-olefins, unreacted ethylene, aluminum alkyls of the same carbon number distribution as the product olefins, and polymer. The effluent is flashed to remove ethylene, filtered to remove polyethylene, and treated to reduce the aluminum alkyls in the stream. In the original plant operation, these aluminum alkyls were not removed, resulting in the formation of paraffins (- 1.4%) when the reactor effluent was treated with caustic to kill the catalyst. In the new plant, however, it is likely that these aluminum alkyls are transalkylated with ethylene by adding a catalyst such as 60 ppm of a nickel compound, eg, nickel octanoate (6). The new plant contains a caustic wash section and the product olefins still contain some paraffins ( 0.5%). After treatment with caustic, cmde olefins are sent to a water wash to remove sodium and aluminum salts. 

SASOL. SASOL, South Africa, has constmcted a plant to recover 50,000 tons each of 1-pentene and 1-hexene by extractive distillation from Fischer-Tropsch hydrocarbons produced from coal-based synthesis gas. The company is marketing both products primarily as comonomers for LLDPE and HDPE (see Olefin polymers). Although there is still no developed market for 1-pentene in the mid-1990s, the 1-hexene market is well estabhshed. The Fischer-Tropsch technology produces a geometric carbon-number distribution of various odd and even, linear, branched, and alpha and internal olefins however, with additional investment, other odd and even carbon numbers can also be recovered. The Fischer-Tropsch plants were originally constmcted to produce gasoline and other hydrocarbon fuels to fill the lack of petroleum resources in South Africa. 

The conversion of fatty alcohols is approximately 99%. The reaction product is then condensed and sent to a distillation column to remove water and high boilers. Typically, a-olefin carbon-number distribution is controlled by the alcohol composition of the reactor feed. The process is currentiy used to produce a-olefins from fatty alcohols. A typical product composition is at <5%, at 50—70%, C g at 30—50%, C2Q at <2%,... 

When determining the product selectivities, all compounds of equal carbon numbers (paraffines, olefins, isomers, and oxygen compounds) were summarized to one product fraction. The chain growth probability was determined by the Anderson-Schulz-Flory (ASF) distribution ... 

Acidic isomerization of the C5-C6 naphtha and some heavy alcohols from the aqueous product refinery (not shown in Figure 18.5) produced a reasonable-quality olefinic motor gasoline (Table 18.10). The octane value varied depending on the carbon number distribution of the feed, which could result in a product with an octane number up to ten units higher. 

Iron. Fe-Cu-K data are shown in Table 2. Carbon number distribution (Flory plots) and a-olefin/paraffin data are shown in Figures 3 and 4, respectively. As on Ru catalysts, the Flory plot is curved and the a-olefin/paraffin ratio decreases to zero as carbon number increases. The experimental conditions are different for the Ru and Fe systems therefore, we cannot make direct comparisons. Such comparisons will be made in a later publication (14). However, we comment on three important findings. First, C2 and C3 hydrocarbons fall close to the Flory curve (Figure 3) for Fe, in clear contrast with the results on Ru. This suggests that the high rate of ethylene readsorption that leads to low C2 concentrations on Ru (7,8) does not occur on Fe. Secondly, both a-olefins and / -olefins persist at higher carbon numbers than on Ru the Flory plot for Fe shows a more pronounced curvature and the asymptotic value of a is reached at higher carbon number than on Ru. Finally, Ru catalysts produce about 40 wt% C20+ product whereas the Fe... 

The trends in carbon number distribution and in a-olefin/paraffin ratio on Ru, Fe, and Co, three very different catalytic surfaces, are remarkably similar. All catalysts show a curved Flory plot and an a-olefin/paraffin ratio that decreases with increasing carbon number until only paraffins are observed at high carbon numbers. In each case, diffusion-enhanced olefin readsorption accounts for such trends. Its contribution depends on the catalytic surface, its physical structure, and reaction conditions. 

CO reactants and the H2O product of the synthesis step inhibit many of these secondary reactions. As a result, their rates are often higher near the reactor inlet, near the exit of high conversion reactors, and within transport-limited pellets. On the other hand, larger olefins that are selectively retained within transport-limited pellets preferentially react in secondary steps, whether these merely reverse chain termination or lead to products not usually formed in the FT synthesis. In later sections, we discuss the effects of olefin hydrogenation, oligomerization, and acid-type cracking on the carbon number distribution and on the functionality of Fischer-Tropsch synthesis products. We also show the dramatic effects of CO depletion and of low water concentrations on the rate and selectivity of secondary reactions during FT synthesis. 

Carbon number distributions are similar on all Co catalysts. As on Ru catalysts, termination probabilities decrease with increasing chain size, leading to non-Flory product distributions. The modest effects of support and dispersion on product molecular weight and C5+ selectivity (Table III) reflect differences in readsorption site density and in support pore structure (4,5,14,40,41), which control the contributions of olefin readsorption to chain growth. Carbon number distributions obey Flory kinetics for C30+ hydrocarbons the chain growth probability reaches a constant value (a ) as olefins disappear from the product stream. This constant value reflects the intrinsic probability of chain termination to paraffins by hydrogen addition it is independent of support and metal dispersion in the crystallite size range studied. 

Ethylene and propylene intrinsic readsorption rates exceed those of larger olefins (4). This leads to the low apparent chain termination probabilities for C2 and C3 molecules shown in Fig. 8. These lower /8t values reflect a more effective reversal of chain termination to olefins (/3o) for these smaller but more reactive molecules (Figs. 9 and 10). The high rate of ethylene and propylene readsorption accounts for their appearance below the rest of the distribution in carbon number distribution plots on both Co and Ru catalysts (Fig. 5). 

Higher intrapellet residence times increase the contribution of chain initiation by a-olefins to chain growth pathways. This intrapellet delay, caused by the slow diffusion of large hydrocarbons, leads to non-Flory carbon number distributions and to increasingly paraffinic long hydrocarbon chains during FT synthesis. But intrapellet residence time also depends on the effective diameter and on the physical structure (porosity and tortuosity) of the support pellets. The severity of transport restrictions and the probability that a-olefins initiate a surface chain as they diffuse out of a pellet also de-... 

Experimental number distributions (Fig. 16) and chain termination probabilities (/3 and jSn) (Fig. 17) on Co catalysts at low values of bed residence time (<2 s, <10% CO conversion) are accurately described by the model. We reported previously a similar agreement on Ru catalysts (4). The model quantitatively describes the observed curvature of carbon number distribution plots (Fig. 16) and also the constant values of j3/y and the decreasing values of )3o observed as hydrocarbon size increases (Fig. 17). Such effects arise from the higher intrapellet fugacity and the higher residence time of larger a-olefins within transport-limited pellets. 

The high asymptotic value of C5+ selectivity at large values of occurs on pellets that restrict the removal of reactive a-olefins and allow many readsorption events in the time required for intrapellet olefin removal by diffusion. Yet, transport restrictions within these pellets must not significantly hinder the rate of arrival of CO and H2 reactants to the active sites. Carbon number distributions also obey Flory kinetics for high values of because even the smaller olefins significantly react within a catalyst pellet. 

Our readsorption model shows that carbon number distributions can be accurately described using Flory kinetics as long as olefin readsorption does not occur (/3r = 0), because primary chain termination rate constants are independent of chain size (Fig. 24). The resulting constant value of the chain termination probability equals the sum of the intrinsic rates of chain termination to olefins and paraffins (j8o + Ph)- As a result, FT synthesis products become much lighter than those formed on Co catalysts at our reaction conditions (Fig. 24, jSr = 1.2), where chain termination probabilities are much lower than jS -I- Ph for most hydrocarbon chains. The product distribution for /3r = 12 corresponds to the intermediate olefin readsorption rates experimentally observed on Co/Ti02 catalysts, where intrapellet transport restrictions limit the rate of removal of larger olefins, enhance their secondary chain initiation reactions, and increase the average chain size of FT synthesis products. 

Fig. 24. Effect of enhanced a-olefin readsorption rates on carbon number distributions (simulations experimental/model parameters as in Fig. 16).

Carbon number distribution plots also become linear when olefins readsorb very rapidly (large /3r) or when severe intrapellet transport restrictions (large ) prevent their removal from catalysts pellets before they convert to paraffins during chain termination (Fig. 24, jSr = 100). In this case, chain termination to olefins is totally reversed by fast readsorption, even for light olefins. Chain termination occurs only by hydrogen addition to form paraffins, a step that is not affected by secondary reactions and for which intrinsic kinetics depend only on the nature of the catalytic surface. The product distribution again obeys Flory kinetics, but the constant chain termination probability is given by )8h, instead of po + pH- Clearly, bed and pellet residence times above those required to convert all olefins cannot affect the extent of readsorption or the net chain termination rates and lead to Flory distributions that become independent of bed residence time. 

Diffiisional restrictions increase the effectiveness of olefin interception sites placed within catalyst pellets. Very high olefin hydrogenation turnover rates or site densities within pellets prevent olefin readsorption and lead to Flory distributions of lighter and more paraffinic hydrocarbons. Identical results can be obtained by introducing a double-bond isomerization function into FT catalyst pellets because internal olefins, like paraffins, are much less reactive than a-olefins in chain initiation reactions. However, light paraffins and internal olefins are not particularly useful end-products in many applications of FT synthesis. Yet, similar concepts can be used to intercept reactive olefins and convert them into more useful products (e.g., alcohols) and to shift the carbon number distribution into a more useful range. In the next section, olefin readsorption model simulations are used to explore these options in the control of FT synthesis selectivity. 

In contrast with readsorption reactions, which broaden the carbon number distribution of FT synthesis products, cracking reactions narrow such distributions, within the constraints imposed by the random nature of C—C bond cleavage in carbenium ion reactions of large olefins. Cracking of n-paraffins can also occur on intrapellet acid sites, but acid-catalyzed paraffin reactions are much slower than those of corresponding olefins of equal size. As in all secondary reactions, cracking sites are used most efficiently when... 

Fig. 27. Effect of diffusion-enhanced a-olefin cracking catalytic function on carbon number distribution (simulations experimental/model parameters as in Fig. 15, 10% CO conversion). (A) FT synthesis without cracking function (B) with intrapellet cracking function, jS = 1.2 (C) with extra pellet cracking function, jS = 1.2. (a) Carbon selectivity vs. carbon number (b) Flory plots.

Fig. 30. The effect of secondary reactions on FT synthesis selectivity. Composite Fe-Mn/ Y-zeolite and Fe-Mn/SiOi catalysts (501 K, 2500 kPa, H2/CO = 2.1, 45-47%.5% CO conversion). (a) Carbon number distribution (b) a-olefin to n-paraffin ratio.

Diffusion-limited removal of products from catalyst pellets leads to enhanced readsorption and chain initiation by reactive a-olefins. These secondary reactions reverse chain termination steps that form these olefins and lead to heavier products, higher chain growth probabilities, and more paraffinic products. Diffusion-enhanced readsorption of a-olefins accounts for the non-Flory carbon number distributions frequently observed during FT synthesis on Co and Ru catalysts. Diffusion-limited reactant (H2/CO) arrival leads instead to lower selectivity to higher hydrocarbons. Consequently, intermediate levels of transport restrictions lead to highest selectiv-ities to C5+ products. A structural parameter containing the pellet diameter, the average pore size, and the density of metal sites within pellets, determines the severity of transport restrictions and the FT synthesis selectivity on supported Ru and Co catalysts. 

Guczi et al. (305) also find antipathetic structure sensitivity for the CO/H2 reaction over Ru/Al203 and Ru/Si02. Alloying with iron has little effect on TOF but increases selectivity toward olefins. In contrast to some previous reports (128, 306), it has been found (307) that the product carbon number distribution is sensitive to particle size. 

Todic et al. [14] developed a comprehensive micro-kinetic model based on the carbide mechanism that predicts FT product distribution up to carbon number 15. This model explains the non-ASF product distribution using a carbon number dependent olefin formation rate (e term). The rate equations for the olefins and paraffins used in the model are shown in Figure 2. The derivation of the rate equations and physical meaning of the kinetic parameters, as well as their fitted values, can be found in Todic et al. [14]. In the current study, a MATLAB code which uses the Genetic Algorithm Toolbox has been developed, following the method of Todic et al. [14], to estimate the kinetic model parameters. In order to validate our code, model output from Todic et al. [14] was used as the input data to our code, and the kinetic parameter values were back-calculated and compared to the values fi om [14], as shown in Table 1. The model has 19 kinetic parameters that are to be estimated. The objective function to be minimized was defined as... 

---