Conversion of Glycerol to 1,3-Propanediol via Selective

May 31, 2003 - All of these steps are experimentally examined in the current work to verify the feasibility of the new conversion approach. Before thi...
52 downloads 15 Views 159KB Size
Ind. Eng. Chem. Res. 2003, 42, 2913-2923

2913

Conversion of Glycerol to 1,3-Propanediol via Selective Dehydroxylation Keyi Wang, Martin C. Hawley,* and Scott J. DeAthos Department of Chemical Engineering and Materials Science, Michigan State University, East Lansing, Michigan 48824

This paper presents a new approach to the production of 1,3-propanediol from glycerol via selective dehydroxylation. The idea is to selectively transform the middle hydroxyl group of glycerol into a tosyloxyl group and then remove the transformed group by catalytic hydrogenolysis. With this new approach, the conversion of glycerol to 1,3-propanediol is completed in three steps, namely, acetalization, tosylation, and detosyloxylation. All of these steps are experimentally examined in the current work to verify the feasibility of the new conversion approach. Before this work, the hydrogenolysis of tosylates had never been reported to have been performed catalytically with H2 as the reducing reagent. This work shows that this reaction can be effected, provided that the tosylate of concern contains a hydroxyl group next to its tosyloxyl group. The mechanism of this reaction is also established. Introduction 1,3-Propanediol (1,3-PD) is a high-value specialty chemical that is used mainly in specialty polyester fibers, films, and coatings. 1,3-PD does not impart the stiffness of ethylene glycol, but it avoids the floppiness of 1,4-butanediol and 1,6-hexanediol. Polyesters based on 1,3-PD have unique features that are difficult to obtain from other glycols. For instance, poly(trimethylene terephthalate), a new polyester based on 1,3-PD that Shell Chemical recently announced for commercialization, has the elastic recovery of nylon and the chemical resistance of polyester.1 Currently, 1,3-PD is made by the hydration of acrolein to β-hydroxypropionaldehyde, which yields 1,3-PD upon hydrogenation. Acrolein is a hazardous chemical. In addition, the yield from this process is low.2 The problem with yield originates with the first step of the process: because acrolein has a large tendency to polymerize through self-condensation, the hydration reaction has to compete with acrolein self-condensation to produce the desired β-hydroxypropionaldehyde. The low conversion efficiency of the acrolein process, as well as the hazardous nature of acrolein, has spurred a great deal of interest in producing 1,3-PD from other chemical sources, especially glycerol. Che3 invented a one-step catalytic process to convert glycerol to 1,3-PD, as well as 1,2-PD, using a synthesis gas co-feed. However, the glycerol dehydroxylation process enjoying more success and attracting the attention of more investigators is the fermentation process.2,4,5 Nevertheless, this process also has some drawbacks. One of the main drawbacks is its low theoretical yield, owing to the byproducts, such as ethanol, acetate, lactate, butyrate, H2, and CO2, formed in this process. Another main drawback is that the process is substrateinhibited. The bacteria used in the fermentation are generally not able to stand a glycerol concentration above 17%.5 As a result, both the product concentration and the productivity are low. * To whom correspondence should be addressed. E-mail: [email protected]. Fax: 517-432-1105.

This paper presents a new approach to glycerol conversion, together with the experimental work demonstrating its promise. According to this new conversion approach, the middle hydroxyl group of glycerol is selectively transformed into a tosyloxyl group with the help of a group protection technique, and then the transformed group is removed by catalytic hydrogenolysis. The hydrogenolysis of tosylates had never been performed catalytically before this work, but the current work shows that it can be done, if the tosylate contains a hydroxyl group next to its tosyloxyl group, as the one involved in our new conversion approach does. Aside from the low efficiency and hazardous nature of the acrolein process, our interest in producing 1,3PD from glycerol is motivated by additional considerations. The glycerol is a material that can be derived from biomass and can be produced from either animal fats, vegetable oils, or sugars (through sugar fermentation and potentially sugar hydrogenolysis). Considering the inevitability of petroleum depletion in the future, basing the production of a chemical on materials that can be derived from biomass has additional attractiveness. Process Concept The new glycerol dehydroxylation approach is illustrated in Figure 1. It consists of three steps, namely, acetalization, tosylation, and detosyloxylation The idea is to selectively transform the second hydroxyl group of glycerol into a tosyloxyl group (tosylation) and then remove the tosyloxyl group by catalytic hydrogenolysis (detosyloxylation). Compared to the hydroxyl group, the tosyloxyl group is a better leaving group and is easier to replace with a hydride ion. In the remainder of this section, a discussion of each of the steps involved in the new conversion approach is provided. The first step in the conversion of glycerol to 1,3-PD is to acetalize the glycerol with benzaldehyde. The purpose is to protect the first and third hydroxyl groups of glycerol, so that only the middle one can be tosylated in the second step and subsequently removed in the third step. The condensation between glycerol and

10.1021/ie020754h CCC: $25.00 © 2003 American Chemical Society Published on Web 05/31/2003

2914

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003

Figure 1. Illustrations of the new approach to glycerol conversion.

benzaldehyde is an equilibrium reaction, but it can be driven to completion by removing the water formed in the reaction. The difficulty with this step is that, in addition to the desired 1,3-product (5-hydroxyl-2-phenyl1,3-dioxane or HPD), the undesired 1,2-product (4hydroxylmethy-2-phenyl-1,3-dioxolane or HMPD) is also formed in the reaction. These products need to be separated. The separated 1,2-product can be returned to the acetalization reactor, where it can either be converted into the 1,3-product or help shift the reaction toward the 1,3-product, to take advantage of the equilibrium nature of the acetalization reaction. Glycerol has been successfully acetalized with benzaldehyde by several researchers.6-9 The reaction actually yielded four products: cis- and trans-HPD and cisand trans-HMPD. Separation of the resulting HPD (1,3product) from HMPD (1,2-product) presented little difficulty.6,7 It was found that the solubility of cis-HPD in a benzene-ligroin mixture is considerably different from those of the other three products, so separation of cis-HPD can be accomplished by crystallizing it from this mixture. Shower and Darley10 reported that cisHPD is the more stable diastereomer of HPD.

The reported data on the equilibrium ratio between the 1,2- and 1,3-products varies widely depending on the investigator and the conditions under which the data were obtained. Hill et al.6 showed that the 1,2product is favored over the 1,3-product by a ratio of 5.7:1 in the absence of solvent. Data from Baggett et al.8 also favor the 1,2-product, but only by a ratio of 1.9:1. The data from Jochims and Kobayshi11 are even more different. They show that the 1,3-product is favored over the 1,2-product by a ratio of 1.5:1. Another example that the reaction favors the 1,3-product was given by Juaristi and Antiunez,12 who obtained a 47:30:23 mixture of HMPD, cis-HPD, and trans-HPD from the reaction between glycerol and benzaldehyde. Upon continued exposure to acid, this ratio changed to 22:34:44. Despite the inconsistency in the reported data, the above work clearly shows that the 1,3-product HPD can be quantitatively produced from the reaction between glycerol and benzaldehyde and that the 1,2- and 1,3-products, or HPMD and HPD, respectively, are readily interconvertible under the acetalization conditions used. Aside from benzaldehyde, many other carbonyl compounds could also be used as the hydroxyl group

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003 2915

protection reagent. An extensive review of the condensation reactions between glycerol and various aldehydes and ketones can be found in Showler and Darley.10 Generally speaking, reactions between glycerol and aldehydes give more 1,3-product than reactions between glycerol and ketones.13 For our purposes, it is apparent that aldehydes are preferred to ketones. Aside from the product yield, the ease with which the 1,3-product can be separated from the 1,2-product is another important factor that needs to be considered in the selection of the group protection reagent. In terms of the latter aspect, benzaldehyde is superior to aliphatic aldehydes. The products from aliphatic aldehydes are generally high-boiling-point oils, and the differences between the boiling points of the 1,2- and 1,3-products are usually small.14,15 For example, the condensation between glycerol and acetaldehyde gives four products: cis- and trans-5-hydroxyl-2-methyl-1,3-dioxane (1,3product) and cis- and trans-4-hydroxylmethyl-2-methyl1,3-dioxolanes (1,2-product). As documented in Beilstein, the boiling points of these products are 73, 100, 86, and 94 °C, respectively, at 18 Torr. In contrast, the 1,2- and 1,3-products from benzaldehyde have considerably different solubilities in benzene-ligroin mixture. The 1,3product can be readily crystallized from the benzeneligroin solution. This is why benzaldehyde was chosen over other potential protection reagents. The second step of the conversion is tosylation of the unprotected hydroxyl group of the acetalized glycerol, or HPD, so as to transform it into a good leaving group (a tosyloxyl group in this case). The tosylation of HPD has also been achieved by previous investigators. For example, Juaristi and Antiunez12 tosylated both cis- and trans-HPD in pyridine and obtained a 2-phenyl-5-tosyl1,3-dioxane (PTD) yield as high as 98% in both reactions. A high PTD yield (94%) was also achieved by Hessel et. al.16 The above work suggests that the tosylation of HPD is a fairly straightforward process. The final step of the conversion is a detosyloxylation reaction preceeded or followed by a hydrolysis reaction. The detosyloxylation reaction removes the tosylated middle hydroxyl group, while the hydrolysis reaction removes the protection on the first and third hydroxyl groups. This last step yields the conversion target, 1,3PD. It also regenerates the group protection reagent benzaldehyde, which can be recycled back to the acetalization reactor for reuse in the first-step conversion. As shown in Figure 1, there are two potential approaches to accomplish the last-step conversion. Either the hydroxyl group protection can be removed first, followed by the tosyloxyl group, or the tosyloxyl group can be removed first, followed by the hydroxyl group protection. If the detosyloxylation reaction is carried out first to remove the tosyloxyl group, 2-phenyl-1,3-dioxane results as an intermediate. Subsequent hydrolysis of this intermediate product is expected to be easy, because no functional groups other than the ether linkages should be affected. If the hydrolysis reaction is carried out first to remove the hydroxyl group protection, 2-tosyloxy-1,3-propanediol (TPD) is generated as the intermediate product. Caution needs to be exercised in performing this reaction to prevent the potential attack of solvents on the tosyloxyl group of the starting PTD and the resulting TPD. Although, according to Flowers,17 tosylates are exceedingly stable under the acid hydrolysis conditions, and although many successful examples of tosylate hydrolyses have been documented

by Tipson,18 our experiments show that the tosyloxyl group is subject to the attack of protic solvents at temperatures above 60 °C. Overall, however, the hydrolysis reaction is simple and easy to perform. The detosyloxylation reaction shown in Figure 1 is basically a hydrogenolysis reaction. According to the proposed conversion approach, this reaction is to be done with molecular hydrogen in the presence of a transition metal catalyst. This reaction is expected to be the most difficult of all of the reactions involved in the new conversion approach, because the tosylate has never before been reported to have been hydrogenolyzed catalytically with H2 as the reducing reagent. At the current time, the hydrogenolysis of tosylates is generally effected with a lithium hydride, either LiAlH4 or LiHBEt3, but these reagents are too expensive for use on an industrial scale. Therefore, the feasibility of catalytically hydrogenolyzing the tosylate is the focus of the current research. Experimental Section 1. Acetalization. The acetalization of glycerol with benzaldehyde was conducted in benzene. A standard experimental setup as described in Casey et al.19 was used for this reaction. This setup includes a roundbottomed reaction flask, a condenser, and a Dean-Stark trap. By using a Dean-Stark trap, the water formed in the reaction could be boiled off from the reaction flask as an azeotrope with benzene (boiling point ) 65 °C), and the reaction could be driven to completion. In a typical experiment, 100 g of glycerol, 120 g of benzaldehyde (ca. 6% excess), and 300 mL of benzene, together with 1 g of p-toluenesulfonic acid catalyst, were placed in the reaction flask. The reaction was initiated by bringing the reaction solution to a boiling state, and the progress of the reaction was monitored by the volume of the water formed in the reaction. Once the reaction was complete, the benzene solvent was boiled off, leaving only the product in the reaction flask. The product of this reaction was a mixture of cis- and trans-5-hydroxy-2-phenyl-1,3-dioxane (HPD, the 1,3product) and cis- and trans-4-hydroxy-methyl-2-phenyl1,3-dioxolane (HMPD, the 1,2-product). A small amount of benzaldehyde normally also was present in the product, as a result of the excess benzaldehyde used in the reaction. Isolation of the desired HPD from the product mixture was accomplished by crystallizing it from a 1:1 benzene-ligroin solution using a procedure taken from Hill et al.6 Analysis of the raw reaction product was performed by GC. Characterization of the purified HPD was done with 1H and 13C NMR spectroscopies. The GC column used was a DB-5 nonpolar column purchased from J&W Scientific Company. The carrier gas was hydrogen (at a flow rate of 1 mL/min). The oven temperature was programmed so that it was held at 150 °C for 1 min, was then increased at a rate of 40 °C/min to 320 °C, and was then held at this temperature for 3 min. The retention times for the relevant compounds are 1.30 min for benzene, 1.42 min for benzaldehyde, 3.36 min for cis-HPD, 3.48 min for both cis- and trans-HMPD, and 3.83 min for trans-HPD. Typically, the chromatograph exhibited three product peaks, those for cis-HPD, transHPD, and HMPD. cis- and trans-HMPD could not be separated with the GC column that we used. 2. Tosylation. Tosylation of 5-hydroxy-2-phenyl-1,3dioxane (HPD) was carried out in pyridine. To start the

2916

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003

experiment, HPD and TsCl were placed in an Erlenmeyer flask. Then, cold dry pyridine was added to the flask. The flask was shaken until both the HPD and TsCl dissolved in the pyridine. Then, the flask was placed in a refrigerator (ca. 5 °C) to allow the reaction to continue for about 12 h. Typically, 5-10% excess TsCl was used in this reaction, and for each gram of HPD, about 20 mL of pyridine was needed. The progress of this reaction was monitored by the formation of needle-shaped crystals. These crystals are a pyridine-hydrochloride complex. When no more crystals were being formed, the reaction was considered complete. At this time, the flask was removed from the refrigerator. The contents of the flask were poured with stirring into a beaker containing about 6 g of ice for each milliliter of pyridine. The tosylate product (2-phenyl-5tosyl-1,3-dioxane or PTD) precipitated immediately. Stirring was continued until all of the ice melted. Then, the solution was filtered with a Buchner funnel under vacuum, and the crystals were washed with water three times. Finally, the crystals were dried in a desiccator under vacuum. After this procedure, PTD was obtained as white waxlike crystals. Characterization of the PTD was accomplished using 1H and 13C NMR spectroscopies. 3. Hydrolysis. Hydrolysis of PTD was performed in both water and methanol. To carry out the reaction in water, PTD from the tosylation reaction was placed in a glass vial equipped with a magnetic stirring bar. Then, a small amount of p-toluenesulfonic acid (TsOH) and a large amount of water were added to the glass vial. Because PTD is not soluble in water, it floated on top of the water at the beginning of the reaction. The glass vial was heated in a water or sand bath with constant stirring of the reaction solution. After 4-6 h of heating and stirring, the PTD floating on the water disappeared, and benzaldehyde was seen as yellow oil particles at the bottom of the glass vial. As the reaction mixture cooled and stood undisturbed, benzaldehyde separated from the aqueous phase. The benzaldehyde phase was removed from the glass vial. The hydrolysis product 2-tosyl-1,3-propanediol (TPD) is believed to have been mostly in the aqueous phase at this point. To carry out the hydrolysis reaction in methanol, PTD was again placed in a glass vial. Then, a methanol solution of TsOH was added to the vial. The glass vial was simply placed in an oven at 40 °C. After 2 h, the reaction was complete. The benzaldehyde formed in the reaction did not separate from the methanol phase. To isolate PTD, the reaction product was first extracted with hexane to remove benzaldehyde, and it was then dried at a temperature below 40 °C to remove the methanol solvent. The thus-obtained PTD was redissolved in dioxane and then transferred to the hydrogenolysis reactor. Of the above two procedures, the latter was more successful. The hydrolysis reaction in water is usually carried out at a temperature above 80 °C. At this temperature, tosylates are subject to attack by protic solvents, as discovered in this work. The hydrolysis reaction in methanol is carried out at a temperature of 40 °C. At this temperature, solvent attack on the tosyloxyl group is relatively insignificant. The product of the hydrolysis of PTD (TPD) was not characterized directly, but was characterized indirectly through the 1,3-PD obtained from the subsequent hydrogenolysis of this product.

4. Hydrogenolysis. Several tosylates were subjected to hydrogenolysis conditions. The reaction was carried out under hydrogen pressure, with either Raney Ni, Ni on kieselguhr, or Ru on carbon as the catalyst. The reactor was a stainless steel batch reactor. The solvents used included water and dioxane. All hydrogenolysis products were characterized by GC, with the same column and method as in the acetalization experiments discussed earlier. The cyclohexanol and cyclohexanone from the hydrogenolysis of 2-tosyl-cyclohexanol, however, were additionally characterized with the GC-MS. Finally, all of the chemicals used in this work, except for the HPD and PTD prepared in this work, were purchased from Aldrich. Results and Discussion 1. Acetalization. Acetalization with benzaldehyde protects the first and third hydroxyl groups of glycerol from being tosylated and removed in the subsequent reactions. This reaction has two important features: (a) it is an equilibrium reaction, but it can be driven to completion, and (b) it is nonselective, meaning that, in addition to the desired 1,3-product HPD, the undesired 1,2-product HMPD is also produced and has to be separated from HPD and returned to the acetalization reactor. Because of the latter feature, the percentage of HDP in the acetalization product and the ease of separation of HPD from HMPD are of particular concern. Three acetalization experiments were performed. Two were done in benzene, and one was done in a 1:1 benzene-ligroin mixture. In the first experiment, 50 g of glycerol and 60 g of benzaldehyde (ca. 5% excess) were condensed in 300 mL of benzene. The reaction required about 3 h to complete. GC analysis of the raw reaction product revealed three product peaks. Of these three peaks, two were determined to result from HPD (cisand trans-HPD) and one from HMPD. The two diastereomers of HMPD could not be separated with the GC column that we used. To separate it from the product mixture, HPD was crystallized from a 1:1 benzene-ligroin mixture according to the procedure described by Hill et al.6 The crystallized product was characterized by 1H and 13C NMR spectroscopies. The obtained spectra were found to be consistent with the structure of HPD and with spectra documented for this compound in the literature.12 Measurement of the melting point of this crystallized product (75-80 °C) further identified it as the cis diastereomer of HPD. The previously reported melting points are 37 °C for HMDP, 65 °C for trans-HPD, and 83 °C for cis-HPD. The crystallized cis-HPD was not 100% pure. It contained a small amount of trans-HPD, but very little HMPD. HMPD is believed to not crystallize with cis-HPD, so that it could be washed away from the crystallized HPD. trans-HPD might crystallize with cis-HPD, although to a much lower extent. Both cis- and trans-HPD are desired from the acetalization reaction. From the first experiment, about 10 g of cis-HPD was obtained after purification. The yield was calculated as 25%. Before purification, the yield of cis-HPD was estimated to be about 42% on the basis of the GC peak sizes. The ratio of cis-HPD to trans-HPD to HMPD was 42:33:25. In the second experiment, 100 g of glycerol was condensed with 120 g of benzaldehyde in 300 mL of benzene. The (cis-) HPD yield from this experiment was similar to that of the first experiment.

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003 2917

From these results, it is clear that HPD can be obtained from the acetalization reaction in high yield. The procedure for separating HPD from the product mixture is fairly simple and easy to carry out. However, the undesired HMDP and the unrecovered HPD in combination still account for about 75% of the reaction product. These products need to be recycled back to the acetalization reactor, if an industrial process is to be designed. After crystallization, the undesired HMDP and the unrecovered HPD remained in the benzene-ligroin mixture. If acetalization is subsequently carried out in the same solvent, recycling of this residue will be straightforward. This consideration motivated a third experiment, which was run in a 1:1 benzene-ligroin mixture to explore the feasibility of carrying out the acetalization in such a solvent. This experiment turned out to be a success and furnished an HPD yield of 23%. 2. Tosylation. The purpose of tosylation is to transform the unprotected hydroxyl group of HPD into a good leaving group, so that it can be removed in the subsequent reactions. In the current research, this group transformation is accomplished by treating HPD with tosyl chloride (TsCl) in pyridine. Two experiments were performed. In one experiment, 2 g of HPD was treated with 2.4 g of TsCl (ca. 12% excess), which gave 3.1 g of PTD. The yield was calculated as 83%. In the other experiment, 5 g of HPD was treated with 6 g of TsCl (ca. 12% excess), which furnished 8.1 g of PTD. The yield was 87%. The obtained PTD was characterized by 1H and 13C NMR spectroscopies. The obtained spectra are consistent with the structure of PTD and with the data documented in the literature.12 In the literature, a yield as high as 98% has been reported in the tosylation of HPD.12 The lower yield in our experiments might result from the unpurified TsCl and the undried pyridine used in this experiment, which are believed to affect the yield. However, no additional effort was made in the current research to maximize the product yield. According to the results of previous research, as well as our own experiments, we are convinced that a higher yield can be achieved in the tosylation of HPD. The purpose of the current experiments was mainly to provide the PTD needed for the subsequent detosyloxylation experiments. 3. Detosyloxylation. Detosyloxylation, or removal of the tosyloxyl group, is the focus of this work. This reaction has never been done with molecular hydrogen as the reducing reagent. In this work, we attempt to develop such a hydrogenolysis process to remove the tosyloxyl group from either TPD or PTD and to create a pathway from PTD to 1,3-PD. As discussed previously, the conversion of PTD to 1,3PD can potentially be approached in two different ways. One approach is to first remove the hydroxyl group protection and then remove the tosyloxyl group. The other is to first remove the tosyloxyl group and then remove the hydroxyl group protection. Both approaches were explored in this work. The second approach proved to be unsuccessful. The PTD was found to be inactive under the hydrogenolysis conditions. The experiment was carried out in dioxane with Ni on kieselguhr as the catalyst, and the reaction was run under 3 MPa H2 pressure. The temperature was kept at 120 °C for 2 h, at 130 °C for 2 h, at 140 °C

Figure 2. Reaction pathway of TCH hydrogenolysis.

for 10 h, and then at 150 °C for another 10 h. No product was found at the end of the experiment. The initial attempt to convert PTD to 1,3-PD via the first approach did furnish 1,3-PD. However, the yield was low (less than 15%). To start this experiment, the PTD was hydrolyzed at 80 °C in 0.01 N aqueous p-toluenesulfonic acid solution. This reaction normally takes about 4-5 h. After the reaction, the benzaldehyde released from the reaction, which separated itself from the aqueous phase, was removed. The aqueous phase, which was believed to contain most of the TPD resulting from the hydrolysis, was neutralized with NaOH and transferred to the hydrogenolysis reactor to remove the tosyloxyl group. Three catalysts, including Raney Ni, Ni on kieselguhr, and Ru on carbon, were used in the hydrogenolysis, and none of them provided a 1,3-PD yield higher than 15%. Despite the low 1,3-PD yield obtained in this experiment, it shows that TPD, in contrast to PTD, is reactive under hydrogenolysis conditions. Comparing the structures of PTD and TPD, the tosyloxyl group of PTD is isolated, whereas the tosyloxyl group of TPD is next to a hydroxyl group. The presence of this adjacent hydroxyl group must have somehow facilitated the removal of the tosyloxyl group of TPD. To confirm the above speculation, experiments were carried out to hydrogenolyze 6-tosyloxyhexanol (TH) and 2-tosyloxycyclohexanol (TCH). Like PTD, TH has an isolated tosyloxyl group, whereas like TPD, the tosyloxyl group of TCH is next to a hydroxyl group. As expected, TH was found to be inactive under the hydrogenolysis conditions. TCH, on the other hand, was successfully hydrogenolyzed at temperatures above 140 °C. Below 140 °C, the reaction was slow. The hydrogenolysis reactions of both TH and TCH were conducted in dioxane. The hydrogenolysis of TCH was conducted with both Ni on kieselguhr and Ru on carbon. The hydrogenolysis of TH, however, was conducted only with Ni on kieselguhr. The results of these experiments confirm our speculation that the presence of a hydroxyl group next to the tosyloxyl group facilitates removal of the latter via hydrogenolysis. Hydrogenolysis of TCH yields two products. One is cyclohexanol, and the other is cyclohexanone, as found in our GC-MS analysis. With Ni on kieselguhr as the catalyst, the ratio of cyclohexanol to cyclohexenone in the product is about 1:1; with Ru on carbon as the catalyst, the ratio is about 3:1. No other products were found. Judging from the observed reaction products, the hydrogenolysis of TCH might occur as shown in Figure 2, with cyclohexanone as the precursor to cyclohexanol. The function of the transition metal catalyst in this reaction is not limited to catalyzing the hydrogenation of cyclohexanone to cyclohexanol in the last step. The

2918

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003

catalyst must be involved in the initial detosyloxylation step as well, because, in the absence of the transition metal catalyst, TCH was found to be inactive at 140 °C. The assistance of the adjacent hydroxyl group to detosyloxylation can be understood on the basis of the assumption that the initial detosyloxylation step in the reaction scheme in Figure 2 is thermodynamically unfavorable. The presence of the adjacent hydroxyl group stabilizes the initial detosyloxylation product through a subsequent tautomerization reaction as shown in Figure 2. In the absence of such a hydroxyl group, as in the cases of PTD and TH, the subsequent tautomerization is impossible, and thus, the initial detosyloxylation product is not stabilized. This explains why PTD and TH are inactive under hydrogenolysis conditions. To understand why the 1,3-PD yield was so low in our initial attempt to convert PTD to 1,3-PD, further experiments were carried out to hydrogenolyze TCH in 90% aqueous methanol. To our surprise, this reaction yielded neither cyclohexanol nor cyclohexanone. The reaction did provide two products, but they are believed to arise from the reactions of TCH with water and with methanol. Further experiments showed that the reactions of TCH with water and methanol can both be effected simply by heating TCH in these solvents over 60 °C. At 80 °C, these reactions become fairly fast. It has been suggested in the literature that the tosylate is stable under hydrolysis conditions.17 This statement appears not to be true for TCH, and it is probably not true for PTD and TPD either. In the hydrolysis of PTD, it was observed that the reaction solution became substantially more acidic at the end of the reaction than at the beginning of the reaction. The increase in the acidity of the reaction solution is a strong sign that the attack of water on the tosylate has occurred. The reaction of water (and methanol) with tosylates releases TsOH, which increases the acidity of the reaction solution. On the basis of the above study on the TCH hydrogenolysis reaction, a new procedure was developed to convert PTD to 1,3-PD. This procedure starts with a hydrolysis reaction in methanol. PTD has a much better solubility in methanol than in water. Therefore, the reaction can be carried out at a temperature as low as 40 °C, while still giving a reasonable reaction rate (normally completed in 2 h). No acidity increase of the reaction solution was observed at the end of the reaction, which implies that the solvent attack on the tosyloxyl group is insignificant at 40 °C. Hydrolysis of PTD yields TPD and benzaldehyde (in the form of acetal with methanol). To separate the benzaldehyde from the TPD, the raw product was extracted with hexane. Then, the reaction solution was dried at room temperature to remove methanol. The dried product was mostly TPD. To hydrogenolyze TPD, it was dissolved in dioxane, and the mixture was then heated to 140 °C in the presence of Ni on kieselguhr. Dioxane is an aprotic solvent, and it does not attack tosylate. The first experiment performed with the new procedure started with 213 mg of PTD. At the end of hydrogenolysis, 27 mg of 1,3-PD was obtained in the product. The yield was calculated as 56%. In addition to 1,3-PD, the hydrogenolysis of TPD also gives npropanol (24%). This is understandable, because the hydrogenolysis of TPD involves β-hydropropionaldehyde

Figure 3. Reaction pathway of TPD hydrogenolysis.

(HPA) as an intermediate, which can also undergo dehydration to give n-propanol, as shown in Figure 3. From Figure 3, one can see that the selectivity of TPD hydrogenolysis is controlled by the relative hydrogenation and dehydration rates of HPA. Increasing the amount of metal catalyst increases the hydrogenation rate. Therefore, the selectivity to 1,3-PD should increase as more catalyst is used. This proves to be true. When the mole ratio of the Ni catalyst to substrate was increased from 1:1 in the first experiment to 2:1 in the second experiment we performed, it was observed that the mole ratio of 1,3-PD to n-propanol in the product increased from 2.3:1 in the first experiment to 4:1 in the second experiment. In the second experiment, 34 mg of 1,3-PD was obtained from 210 mg of PTD. The 1,3-PD yield was 72%. The n-propanol yield was calculated as 18%. No further experiments were conducted in the current work to maximize the 1,3-PD yield. Nevertheless, the above experimental results show that the new procedure is promising for the conversion of PTD to 1,3-PD. In addition to the 1,3-PD yield, the catalytic nature of the TPD hydrogenolysis reaction was another major concern of the current research. A stoichiometric reaction between tosylate and the transition metal has been reported, although in the reported reaction, it was the O-S bond, rather than the C-O bond as in the TPD hydrogenolysis, that was broken.18 Thus, in the current work, the Ni on kieselguhr catalyst used in the second TPD hydrogenolysis experiment was collected after the reaction and was reused in the hydrogenolysis of TCH. It was found that the catalyst was still active, and a turnover number of about 50 was achieved in 24 h at 160 °C and 3 MPa H2 pressure. At this point, there should be no doubt that the hydrogenolysis of TPD in the presence of Ni on kieselguhr is, by nature, a catalytic reaction. 4. Industrial Process Concept. The technical feasibility of the new glycerol dehydroxylation approach presented in Figure 1 was verified on the laboratory scale by the experiments just discussed. On the industrial scale, processes as shown in Figures 4-7 could be designed to produce 1,3-propanediol from glycerol. The acetalization process shown in Figure 4 features a continuous acetalization reactor, a continuous forcedcirculation HPD crystallizer, and a continuous vacuum filter. The reactor is equipped with a water trap operating by the same principle as the Dean-Stark trap used in our benchtop experiments. The purpose is to remove the water formed in the reaction. The design of this process incorporates the following observations made in our acetalization experiments: (a) Acetalization of glycerol with benzaldehyde occurs readily in the presence of acid catalyst and can be driven to completion by using a Dean-Stark trap to remove the water formed

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003 2919

Figure 4. Glycerol dehydroxylation process, step 1: acetalization.

in the reaction. (b) The desired 1,2-product HPD can be readily purified from the product mixture by crystallizing it from benzene-ligroin mixture. The unreacted glycerol and benzaldehyde and the undesired 1,3product HMPD remain in the solution. (c) Acetalization of glycerol with benzaldehyde can be carried out in benzene-ligroin mixture, as well as in benzene only, thus allowing direct crystallization of HPD from the reaction solution and direct recycling of the unreacted glycerol and benzaldehyde and the undesired HMPD back to the reactor, if benzene-ligroin is used as the solvent. According to the design, the HPD-rich product is fed directly into the crystallizer from the reactor. The desired HPD is crystallized from the solution and then separated by filtration. Owing to the nature of the selected crystallizer, some solvent is vaporized in the crystallizer and then condensed in the condenser equipped with the crystallizer. This solvent is collected and used to wash the HPD cake formed on the filter surface. The filter cake discharged from the filter might need further drying. The filtrate, composed of the solvent and the undesired HMPD and the unreacted glycerol and benzaldehyde dissolved in it, is lean in HPD and is recycled back to the reactor. The benzaldehyde used in acetalization is also recycled, but from step 3, the detosyloxylation process. Because of the recycling of HMPD, the process shown in Figure 4 produces no byproduct. Under acetalization condition, HMPD and HPD are readily interconvertible. As a result, the HMPD formed in the reaction can only accumulate to a certain level in the reactor, in equilibrium with the HPD present. Once that level is reached, no additional HMPD will be produced in the reactor. HPD will be the only product of the reaction of glycerol and benzaldehyde, as it is continuously removed from the reactor and process. Therefore, with the process shown in Figure 4, 1 mol of HPD will be expected from each mole of glycerol fed into the reactor, or the HPD yield will be 100% (or close to 100% if possible loss in the filter is considered) under stable conditions. A high

HPD yield is a major advantage of the continuous process shown in Figure 4. The tosylation process shown in Figure 5 is designed strictly on the basis of our laboratory tosylation procedure described in the Experimental Section. It involves three major steps: (a) In the first step, HPD is allowed to react with TsCl in pyridine at low temperature, so that the hydrochloride formed in the reaction can precipitate from the reaction solution as pyridinehydrochloride complex. (b) Following step a, the reaction product is mixed with a large quantity of ice. The purpose is to further lower the temperature and to force the tosylation product PTD to crystallize from the product solution. (c) Finally, PTD crystals are separated by filtration. A high PTD yield was obtained with this procedure in laboratory experiments. As an industrial process, however, this procedure has a major disadvantage: mixing the reaction product with ice in step b creates a difficulty for the recycling of pyridine. On the industrial scale, the pyridine solvent has to be recycled, to minimize the process cost and waste discharge. After PTD removal, the pyridine is in a mixture with water (Figure 5). To reuse pyridine in tosylation, it first has to be separated from water and then thoroughly dried. Tosyl chloride is attacked by the very small quantity of water present in pyridine. To overcome the above difficulty, the alternate process shown in Figure 6 is proposed. This design eliminates ice from the process and attempts to crystallize PTD directly from the reaction medium, thus allowing direct recycling of the medium, together with the unreacted starting material in it. The reaction medium for this process might be pure pyridine or a mixture of pyridine with another solvent, ligroin for example, that helps lower the solubility of PTD in the reaction medium. Pyridine, as a base, is needed in the reaction medium to neutralize the hydrochloride released from the reaction. To remove the pyridine-hydrochloride complex precipitated from the reaction medium, the tosylation reactor in Figure 6 is incorporated with a circulation

2920

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003

Figure 5. Glycerol dehydroxylation process, step 2: tosylation (design 1).

Figure 6. Glycerol dehydroxylation process, step 2: tosylation (design 2).

loop that contains a continuous in-line filter. The reaction product sent to the crystallizer is designed to be free of hydrochloride. The crystallizer in Figure 6 is of the evaporative type, so that the solvent can be partially removed to give a PTD-saturated solution in the crystallizer. The solvent removed from the solution can be used to wash the PTD crystals in the filtration step. At the time being, the solubility data for PTD in pyridine or other relevant solvents are not yet available. It is also left for speculation whether the tosylation reaction can be carried out in a mixture of pyridine with another solvent as smoothly as in pyridine alone. It is clear that further experiments are needed to verify the feasibility of this process. However, if successful, this process has great advantages over the one shown in Figure 5, in terms of both processing cost and waste discharge. As for the acetalization process in Figure 4,

this process can potentially furnish a product yield of 100%, under ideal conditions. The detosyloxylation process shown in Figure 7 is designed on the basis of the discovery made in this research: the detosyloxylation of PTD can be accomplished by a hydrolysis (or methnolysis, more accurately) reaction carried out in methanol at moderate temperature, followed by a catalytic hydrogenolysis reaction carried out in dioxane. This design features a hydrolysis reactor, an extraction column to remove benzaldehyde from the hydrolysis product solution, a distillation column to separate the extracted benzaldehyde from the hexane extractant, a thin-film evaporator/ dryer to remove the methanol solvent from the hydrolysis product TPD, and a hydrogenolysis reactor to finally convert TPD to 1,3-PD. Separation of 1,3-PD from the solvent dioxane and other hydrogenolysis products, including TsOH and propanol, is yet to be added to this

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003 2921

Figure 7. Glycerol dehydroxylation process, step 3: detosyloxylation.

Figure 8. Material flow in glycerol dehydroxylation.

process. According to the design, both the methanol and the hexane solvents used in this process are recycled. The benzaldehyde reclaimed in this process is also recycled, but back to the acetalization process in step 1. Figure 8, obtained by combining and simplifying the processes in Figures 4, 6, and 7, illustrates the material flow in glycerol dehydroxylation. It can be seen from this diagram that the acetalization process, if designed as in Figure 4, produces no product other than HPD, at

nearly 100% yield. Similarly, the tosylation process, if designed as in Figure 6, gives only PTD, at nearly 100% yield. Therefore, the overall yield of the glycerol dehydroxylation process is largely determined by the yield of the detosyloxylation process. In our experiment, a 1,3PD yield as high as 72% was achieved in PTD detosyloxylation, under unoptimized conditions. This yield can be roughly taken as the overall yield of the new glycerol dehydroxylation process. From Figure 8, it can also be seen that the production of 1,3-PD is at the expense of tosyl chloride, gaseous hydrogen, and glycerol. Benzaldehyde is not consumed in the process because of recycling. Assuming an overall process yield of 72%, the production of each pound of 1,3-PD will consume 1.68 lb of glycerol, 3.61 lb of tosyl chloride, and 4.12 mol of hydrogen, which cost about $1.68, $5.60, and $0.12, respectively. The total raw material cost will thus be $7.40 per pound of 1,3-PD, of which 76% results from tosyl chloride. The above costs are calculated on the basis of a unit price of $1.00 per pound for glycerol, $1.55 per pound for tosyl chloride, and $0.03 per mole for hydrogen gas. These prices are from the Chemical Market Report, as well as company sources. It can be seen from the above that the majority of the raw material cost of 1,3-PD production results from tosyl chloride. The price of this chemical is fairly high on the market, as a result of the low demand and thus low production of this chemical. As a solution to reduce the raw material costs, on-site synthesis of tosyl chloride might be a solution. Justification of on-site synthesis of tosyl chloride would require a large-scale glycerol conversion process. The glycerol conversion processes shown in Figures 4, 6, and 7 are mostly continuous and involve no hazardous chemicals; therefore, they are suitable for large-scale production. If the cost of tosyl chloride can be lowered through on-site synthesis, the glycerol dehydroxylation process proposed in this work will become much more attractive.

2922

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003

Theoretically, mesyl chloride might work just as well as tosyl chloride in the new approach to glycerol conversion. Future work might look into the feasibility of using this alternative compound. Because of its lower molecular weight, mesyl chloride is about 40% cheaper than tosyl chloride per mole, although the prices of these two compounds are similar on a per-pound basis. This price difference provides a material cost advantage. Compared to the fermentation process, this new chemical process has the following potential advantages: (a) It can potentially furnish a high product yield. The fermentation process has a limited theoretical yield, which varies with the byproducts it produces. In the ideal case, only acetate would be produced as the byproduct, and the theoretical yield would be 67%. In contrast, the theoretical yield of the new chemical process is not subject to such a limit. (b) The productivity is potentially high. The chemical reactions involved in the new conversion process are all fast reactions, and most of them can be made continuous, as can the separation processes following the reactions, which makes this process very suitable for large-scale production. In contrast, the fermentation process is fairly slow, and its productivity is low. The best result reported so far is 8.1 g L-1 h-1.5 The fermentation process is substrate-inhibited. The bacteria used in the fermentation are generally not able to stand a glycerol concentration above 17%. (c) The separation of the products is potentially easy. The products produced in each step of the new chemical process have dramatically different physical properties. The separation of these products can easily be accomplished with conventional and readily available technologies. In contrast, separation of the fermentation products is difficult. A great deal of research is being dedicated to the separation of 1,3-PD from the glycerol fermentation products. The major disadvantage of the new chemical process is its high raw material cost, which mainly results from consumption of tosyl chloride, a chemical currently produced only on a small scale because of the low demand. A number of possibilities were discussed above to reduce the material cost of the new chemical process, including on-site synthesis of tosyl chloride and substitution of tosyl chloride with the less expensive mesyl chloride. Another possibility is regeneration of tosyl chloride from the p-toluenesulfonic acid (TsOH) produced in the process. If this could be done, then, in theory, no tosyl chloride would be consumed in the conversion process, and the raw material cost would be dramatically reduced. However, a route from p-toluenesulfonic acid to tosyl chloride is yet to be developed.

Conclusions A new glycerol dehydroxylation approach is proposed to convert glycerol to 1,3-propanediol. The idea is to selectively transform the second hydroxyl group of glycerol into a tosyloxyl group (tosylation) and then to remove the transformed group by catalytic hydrogenolysis (detosyloxylation). This new approach involves three steps, all of which are examined experimentally in the current work. The results show that the proposed approach is promising.

Research on this new approach also led to the new finding that the tosyloxyl group of a tosylate can be removed through catalytic hydrogenolysis with H2 as the reducing reagent, provided that this tosylate contains a hydroxyl group next to its tosyloxyl group. The mechanism of this reaction was established. The discovery of this new dehydroxylation process is vital to the success of the proposed new glycerol conversion approach. Acknowledgment This work was generously supported by the Consortium for Plant Biotechnology Research, the Amoco Foundation, the Michigan State University Crop and Food Bioprocessing Center, and the Department of Chemical Engineering at Michigan State University. Literature Cited (1) Stinson, S. C. Fine and Intermediate Chemicals Makers Emphasize New Products and Processes. Chem. Eng. News 1995, 73, 10. (2) Gunzel, B.; Yonsel, S.; Deckwer, W. D. Fermentative production of 1,3-propanediol from glycerol by clostridium butyicum up to a scale of 2 m2. Appl. Microbiol. Biotechnol. 1991, 36, 289-294. (3) Che, T. M. Catalytic Conversion of Glycerol and Synthesis Gas to Propanediol. U.S. Patent 4,642,394, 1987. (4) Heyndrickx, M.; De Vos, P.; Vancanney, M.; De Ley, J. The Fermentation of Glycerol by Clostridium butyricum, LMG 1212t2 and 1213t1 and C. pasteurianum. Appl. Mcrobiol. Biotechnol. 1991, 34, 637. (5) Zeng, A.-P.; Biebl, H.; Schlieker, H.; Deckwer, W.-D. Pathway Analysis of Glycerol Fermentation by Klebsiella pneumoniae: Regulation of Reducing Equivalent Balance and Product Formation. Enzyme Microb. Technol. 1993, 15, 770-773. (6) Hill, H. S.; Whelen, M. S.; Hibbert, H. Studies on the Reactions Relating to Carbonydrates and Polysaccharides. XV. The Isomeric Benzylidene Glycerols. J. Am. Chem. Soc. 1928, 50, 2235. (7) Baggett, N.; Brimacombe, J. S.; Foster, A. B.; Stachey, M.; Weiffen, D. H. Aspects of Stereochemistry. Part IV. Configuration and Some Reactions of the 1,3-O-Benzylideneglycerols (5-Hydroxy-2-Phenyl-1,3-Dioxans). J. Chem. Soc. 1960, 2574. (8) Baggett, N.; Buck, K. W.; Foster, A. B.; Randall, M. H.; Webber, J. M. Aspects of Stereochemistry. Part IV. Absolute Configuration of Benzylidene Derivatives of Some Acylic Polyhydric Alcohols. J. Chem. Soc. 1965, 3394. (9) Szeja, S. Separation of 2-Substituted 5-Hydroxy-1,3-Dioxanes by Esterification in Catalytic Two-Phase System. Pol. J. Chem. 1983, 57, 609. (10) Showler, A. J.; Darley, P. A. Condensation Products of Glycerol with Aldehydes and Ketones: 2-Substituted m-Dioxan5-ols and 1,3-Dioxolane-4-methanols. Chem. Rev. 1967, 67, 427. (11) Jochims, J. C.; Kobayashi, Y. Sterospecific Synthesis of 5-Substituted 1,3-Dioxanes by Application of Dimroth’s Principle. Tetrahedron Lett. 1974, 575. (12) Juaristi, E.; Antunez, S. Conformational Analysis of 5-substituted 1,3-dioxanes. 6. Study of the Attractive Gauche Effect in O-C-C-O Segments. Tetrahedron 1992, 48, 5941. (13) Trosr, B. M. Comprehensive Organic Synthesis; Pergamon Press: New York, 1991; Vol. 6, p 659. (14) Piasecki, A.; Burczyk, B. Acetals and Ethers. Part VI. Synthesis of Selected cis- and trans-2-Alkyl-4-Hydroxylmethyl-1,3Dioxolanes and cis- and trans-2-Alkyl-5-Hydroxy-1,3-Dioxanes. Pol. J. Chem. 1980, 54, 367. (15) Hibbert, H.; Carter, N. M. Studies on the Reactions Relating to Carbohydrates and Polysaccharides. XVIL. Structure of the Isomeric Methylidene Glycerols. J. Am. Chem. Soc 1928, 50, 3120.

Ind. Eng. Chem. Res., Vol. 42, No. 13, 2003 2923 (16) Hessel, L. W.; van Lohuizen, O. E.; Verkade, P. E. GlycerolR- and β-Monoiodohydrin. Rec. Trav. Chim 1954, 73, 842. (17) Flowers, H. M., Protection of the Hydroxyl Group. In The Chemstry of the Hydroxyl Group; Patai, S., Ed.; Interscience: London, 1971; Part 2, pp 1001-1004. (18) Tipson, R. S. Sufonic Esters of Carbohydrate. In Advances in Carbohydrate Chemistry; Academic Press: New York, 1953; Vol. 8, pp 108-215.

(19) Casey, M.; Lenoard, J.; Lygo, B. Advanced Practical Organic Chemistry; Blackie Academic and Professonal: London, 1990; pp 133-134.

Received for review September 24, 2002 Revised manuscript received March 27, 2003 Accepted April 16, 2003 IE020754H