Study of the sequential conversion of citric to ... - ACS Publications

Nanying Bian, Rebecca J.Cunningham, and Maitland Jones, Jr. Department of Chemistry, Princeton University, Princeton, New Jersey 08544. Citric acid ra...
0 downloads 0 Views 1001KB Size
Ind. Eng. Chem. Res. 1994,33, 1989-1996

1989

Study of the Sequential Conversion of Citric to Itaconic to Methacrylic Acid in Near-Critical and Supercritical Water Magnus Carlsson3 Christine Habenicht, Lance C. Kam, and Michael Jerry Antal, Jr.' Hawaii Natural Energy Institute and the Department Manoa, Honolulu, Hawaii 96822

of

Mechanical Engineering, University of Hawaii at

Nanying Bian, Rebecca J. Cunningham, and Maitland Jones, Jr. Department of Chemistry, Princeton University, Princeton, New Jersey 08544

Citric acid rapidly reacts in hot (250 "C), compressed (34.5 MPa) liquid water to form itaconic and citraconic acids with a combined selectivity that exceeds 90%. At higher temperatures (360 "C), in the absence and presence of NaOH, itaconic acid decarboxylates to form methacrylic acid. The yield of methacrylic acid depends on the temperature, pH, and buffer strength of the medium, reaching a maximum of about 70 % (by mole) of the itaconic acid feed. Conditions which favor the production of methacrylic acid also lead to the formation of its hydration product: hydroxyisobutyric acid. Under optimum conditions the combined yield of methacrylic acid and hydroxyisobutyric acid from itaconic acid exceeds 80%. Results are consistent with well-established dehydration and decarboxylation mechanisms.

Introduction Between 200 and 400 million lb of citric acid are produced annually in the USA by fermentation of molasses and other sugars using the microorganism Aspergillus niger (Leeper et al., 1991). A lesser quantity of itaconic acid is manufactured by a similar technology using Aspergillus terreus (Leeper et al., 1991). The recovery of citric acid from its fermentation broth via calcium salt precipitation is a costly, highly complex, sophisticated operation (Bouchard and Merrit, 1979). U.S.D.O.E. estimates the cost of dry citric acid produced from a new plant to be about $0.59/lb, whereas the estimated cost of wet citric acid (in its fermentation broth) from a new plant is about $0.19/lb and from an old plant is about $0.15/lb (Ingham, 1993). Thus the separation of citric acid from its broth triples its cost to the consumer. Itaconic acid is separated from its broth by acidification,clarification, and other purifications, followed by concentration, crystallization, and collection of the acid (Tate, 1981). Thus these two acids depict the principal problem facing all fermentations: how to recover the product from its broth inexpensively. One attack on this problem is to employ the aqueous broth further as a medium for the transformation of the fermentation product into a higher value petrochemical feedstock. If this transformation were to be accomplished in a reactive distillation tower (Doherty and Buzad, 1992), recovery costa could be reduced dramatically. Earlier work in this laboratory examined the utility of near-critical and supercritical water as a reaction medium for the acid-catalyzed conversion of ethanolto ethene (Antal et al., 1987;Xu et al., 1990; 1991), l-propanol and 2-propanol to propene (Ramayya et al., 1987; Narayan and Antal, 1989; 19901, xylose to 2-furaldehyde (Antal et al., 1991), glueose and fructose to 5-(hydroxymethyl)-Z-furaldehyde(Antal et al., 19901,and the dehydration of lactic acid to acrylic acid (Mok et al., 1989). For almost all of these renewable substrates, conditions were identified which result in a favorable yield of the desired product. Moreover, in all cases the possibility exists of avoiding a costly recovery step by employing reactive distillation with water as the reaction t Current

address: TPS, Termiska Processer AB, 5-61182

Nykoping, Sweden.

medium of choice. The goal of this work is to evaluate the utility of near-critical and supercritical water as a medium for the conversion of citric to itaconic acid and itaconic to methacrylicacid. By avoiding costly separations, we hope to realize an economical method for producing a valuable petrochemical from a renewable feedstock. Nevertheless, the loss of two molecules of carbon dioxide and one molecule of water from citric acid results in a 55 % weight loss during the conversion process. Consequently, if the process is to be economical, the reaction chemistry must be quite specific. Because citric acid possesses one hydroxyl and three carboxylgroups, it reacts at elevated temperatures to form a plethora of reactive products that can further participate in a large number of secondary reactions . Acids with a tertiary hydroxyl in the a position are known to decompose to a ketone upon heating (Roberta, 1979),thus citric acid (in the melt phase) decomposes to acetonedicarboxylic acid which further decomposes to acetone (Bouchard and Merritt, 1979). On the other hand, &hydroxy acidsusually dehydrate to an a-p unsaturated acid (sometimes with the loss of carbon dioxide) at elevated temperatures (Hurd, 1929); hence cisltruns-aconitic acid is a major unstable intermediate product of citric acid decomposition in its melt phase. Above 180 "C aconitic acid decomposes to form itaconic and citraconic anhydrides (Bruce, 1943), which revert to their acid counterparts in the presence of water (Bouchard and Merritt, 1979). Prior research (Mok et al., 1989) concerning the dehydration of lactic acid in supercritical water led us to anticipate gains in selectivity if the citric acid reactions were carried out in water. The work on lactic acid suggested the facile formation by intramolecular catalysis of an unstable &lactone intermediate from 8-hydroxy acids, which rearranges at elevated temperatures to form an unsaturated acid. As the hydroxyl group of citric acid is p to two carboxyl groups, we expected it to react in water quickly at elevated temperatures and to form cis1 truns-aconitic acid (see Figure 1). Furthermore, the literature (Arnold et al., 1950)offers ageneral mechanism for the decarboxylationof unsaturated acids, which led us to anticipate the facile conversion of aconitic acid to itaconic acid (see Figure 2) and ita subsequent decarboxylation to methacrylic acid. Because of this prior work

oaaa-~aa5i9412633-19a9$04.5oio 0 1994 American Chemical Society

1990

Ind. Eng. Chem. Res., Vol. 33, No. 8, 1994 COOH

Crotonic acid (CT)

HCQC 1

L

HOOC

Mesaconic acid (MC)

Citraconic acid ( C C )

CCQH COOH Citric acid ( C )

J

Itaconic acid (IC)

ketone (AO) ketic acid (PA) Pyruvic acid (PY) Carbon monodde

- c02 ProDene

Y C O O H

OH

- c02

T

-2 co2-

\

~

OOH

L

COOH

Citramalic acid (CM) Hydroxyisobutyric acid (HIB) Figure 1. Important pathways for decomposition of citric acid in water.

0 II

Methacr)4icacid (MA)

0

Aconitic acid (A)

Itaconic acid (IC)

Methacrylic acid (MA)

Figure 2. Expected mechanism of methacrylic acid formation from aconitic acid (following Arnold et al. (1950)).

Acetoacetic acid

Paraconic acid

COOH HOOC’Y

CH3

W C O O H

Methylsuccinic acid (MS) Vinylacetic acid Figure 3. Structures of various acids relevant to this work. we felt optimistic about the possibility of identifying conditions which would lead to the selective formation of methacrylic acid from either citric or itaconic acid. Nevertheless, the earlier work in aqueous solutions also revealed competing, parasitic pathways which could decrease yields of methacrylic acid at every step of the process. For example, in the lactic acid study (Mok et al., 1989) an acid-catalyzed dehydrationldecarbonylation pathway was found which (in this case) could transform citric acid into acetoacetic acid (see Figure 3). Acetoacetic acid readily decomposes into carbon dioxide and acetone at low temperatures (Loudon, 1988). Moreover,equilibration of the propenedicarboxylic acids is known to be fast in

water solutions at elevated temperatures (Linstead and Mann, 1931;Sakai, 1976);thus itaconic acid may rearrange to citraconic and mesaconic acids (see Figure 1) more rapidly than it decarboxylates to form methacrylic acid. There is no reason to expect either of these two acids to decarboxylate readily and form methacrylic acid. Also, itaconicacid can add water across the double bond, forming citramalic acid. Finally, methacrylic acid is hardly stable at elevated temperatures. Like itaconic acid, methacrylic acid can also add water across its double bond, forming 2-hydroxyisobutyric acid. Methacrylic acid can also decompose by decarboxylation to form propene, which can attack remaining methacrylic acid and form ita propyl ester (Bauer, 1990). Clearly, even in water the selective conversion of citric and itaconic acids to methacrylic acid is a challenging problem in chemical reaction engineering. Apparatus and Experimental Procedures

The principles underlying the design of the plug flow reactors employed in this work have been described in earlier publications (Cutler et al., 1988; Ramayya and Antal, 1989;Xu et al., 1991). The two reactors associated with experiments dated 8/9/93 and before were fabricated from Hastelloy C276 and realized residence times ranging from about 1 to 10 s, and from 20 to over 100 s at

Ind. Eng. Chem. Res., Vol. 33, No. 8,1994 1991 K'kN

o=c

C

-

r

HCI

HO-C--"CN

NaHCO&O L m H z c H I

bo

r

HO-C-"COOH C ' OOH

L m H z c H ,

Figure 4. Synthesis of labeled citric acid.

temperatures as high as 400 "C and pressures up to 34.5 MPa. Recently, these reactors were replaced by a new Hastelloy C276 capillary tube reactor which permits residence times ranging from 1to 15 s, and an Inconel 625 annular flow reactor which enables residence times from 30 to over 400 s. Earlier workers have noted the importance of seasoning (aging) a reactor in order to obtain reproducible results. For example, Lira and McCrackin (1993) observed improved yields of acrylic acid from lactic acid after they suitably aged their Hastelloy reactor. In the context of the work presented here, seasoning did not appear to dramatically affect product yields. Results from the new reactors agreed well with those from the old, wellseasoned reactors. Further details concerning the operation of these reactors were presented in earlier publications (Antal et al., 1987,1990; Ramayya et al., 1987). Samples of the liquid effluent from either reactor were analyzed by HPLC using a 300 X 7.8 mm Biorad Aminex HPX-87H ion exclusion column operated at room temperature with a flow of 0.6 mL/min of 0.01M trifluoroacetic acid solution. Products were detected by a Waters Series R400 RI detector in series with a Hewlett Packard Model 1040A UV-vis diode array detector. Gaseous products were analyzed using a Hewlett-Packard Model 5890 Series I1 GC equipped with a TCD connected in series with a FID, and an Alltech Carbosphere stainless steel packed column (80/100 mesh, 6 f t X 1/8 in). The carrier was an 8% hydrogen in helium mixture flowing at 30 mL/min. The GC furnace was programmed to hold the column a t 35 "C for 4.2 min, followed by a 15 "C/min ramp to 227 "C and a 35 "C/min ramp to 350 "C, which was held for 9.5 min. Daily syringe injections of one to three certified gas standards were used for calibration. Liquid products were identified by a comparison of their retention times and W-vis spectra with those of known standards. These standards were obtained from Sigma Chemical Co. or Aldrich Chemical Co. in the highest available purity (usually 98% or better). No significant impurities in the standards were detected by HPLC analysis. Duplicate injections of at least two samples from the reactor at each reaction condition were subject to analysis. Labeled [l'-l3C]citric acid was employed as a reactant in some experiments. This substrate was synthesized according to the method of Winkel et al. (1989) as follows (see Figure 4). A solution of 8.3 g of NaHS03 in 15 mL of HzO was added slowly to a 100-mL round-bottomed flask containing 11.7 g of diethyl 1,3-acetonedicarboxylate at room temperature. The solution was stirred vigorously with a magnetic stirrer for 30 min. A solution of 5.1 g of

K13CN (99% labeled, purchased from Cambridge Isotope Laboratories) in 11mL of H2O was then added dropwise to the flask, and the mixture was stirred magnetically for 1.5 h. The resulting mixture was slightly basic, so the pH was adjusted to 7 with NaHC03 solution. The contents of the flask were then transferred to a 250-mL separatory funnel for extraction. Approximately 50mL of ethyl ether was added, and the organic layer was then set aside. The aqueous layer was extracted three times with 50 mL of ethyl ether, and the resulting organic layers were combined with the original organic fraction. The combined organic layers were then washed with saturated NaCl solution, dried over MgSOr, and placed in a 250-mL round-bottomed flask. Ethyl ether was removed by evaporation under reduced pressure, yielding a yellow oil, diethyl 3-[13Clcyano-3-hydroxy-l,5-pentanedioate. lH NMR (6,CDCl3, 300 MHz): 1.25 (CH3,m), 2.98 (CHz, dd), 4.24 (OCH2, m). The oily [13C]cyano diester was dissolved in 140 mL of concentrated HCl(12.1 N), and the mixture was refluxed for 17 h. The acidic solvent was removed by evaporation under reduced pressure, yielding a dark brown oil containing white crystals. In order to remove any remaining HC1, three portions of 15 mL of H2O were added to the oil and removed by rotary evaporation. The resulting light brown sludge of crystals and oil was dried overnight in the hood and then dried further in a 66 OC oven for 12 h. The crude, slightly wet crystals were then purified by rinsing with CHCl3 and a few drops of H2O during suction filtration. The resultant solid was then ground with mortar and pestle, yielding 5.4g of fine, yellow-white crystals, mp 151-153 "C which agrees with the literature value of 153 "C (Handbook of Chemistry andPhysics, 1973). 'H NMR (6, acetone-ds, 300 MHz): 2.95 (CH2, dd). For pure, unlabeled citric acid, the following peaks were observed using decoupled 13CNMR (6,acetone-d, 300 MHz): 42.9, 73.4,171.5,174.8. The same peaks appear in the 13CNMR spectrum of the labeled citric acid, with the 6 171.5 peak (Cl') showing significantly increased 13C abundance. To obtain NMR spectra of the reaction products, samples were dried at 68-72 "C and 0.1 Torr in a drying pistol for 2 days. Analysis of authentic mixtures of acids by '3C NMR spectroscopy showed that only citric and itaconic acids were unaffected by this procedure. Methacrylic acid virtually disappeared, and citraconic acid was sharply reduced.

Results and Discussion Tables 1-3 display the results of experiments involving citric acid, itaconicacid, and miscellaneousother reactants. Note that the abbreviations employed in these tables are defined in Figure 1. Product yields are expressed as absolute mole percent (100 X moles of product/mole of reactant fed). In almost all cases, the reactions were conducted a t 34.5 MPa (P,= 1.56) and temperatures

Table 1. Results of Citric Acid Experiments temp

expt

no. reactant 108 106 103 105 3 4 107 6 7

a

0.1 M C 0.5 M C 0.5 M C 0.5 M C 0.5 M C 0.5 M C 0.1 M C 0.5 M C 0.5 M C 0.5 M C

catalyst nil nil nil nil nil nil 2 mM NaOH 10 mM NaOH 10 mM NaOH 10 mM NaOH

res

("C) time (8) 230 230 250 250 280 320 230 280 300 320

247 247 109 156 43 41 247 43 38 44

conv (%) 39 39 43

C02 CO C&

39 36 44 55 60 75 72 100 130 51 49 86 89 100 103 100 121

0 0 1 1 2 7 1 3 5 6

MA IC CC MC CT A 0 AA AD PY HIB C m

Effect of Temperature 0 0 2 8 8 0 0 0 1 1 2 0 0 23 7 1 0 0 1 1 2 0 0 2 8 1 0 1 0 0 3 0 2 0 0 3 4 1 2 1 0 1 2 1 2 0 0 4 1 1 9 3 0 2 3 0 3 0 7 2 6 1 9 1 4 0 6 4 0 2 0 0 3 2 9 1 0 1 1 1 2 0 1 4 6 2 3 4 0 0 3 0 2 2 4 422410 0 6 4 0 2 0 6 3 5 2 5 1 6 0 6 6 0 3

0

9

9 94 0 9 9 0 97 0 95

0

0 8 8 0 9 5 0 8 7 0 92 0 96

sample 9/23/93 e4, e5 9/23/93 c2, c3 8/9/93 84,s5, a6 9/15/93 c2 11/27/92 N2, N4 11/27/92 N16, N18 9/23/93 d 2 4 4 11/27/92 N6, NE, N10 12/28/92 d 1 4 5 11/27/92 n12, n14

1992 Ind. Eng. Chem. Res., Vol. 33, No. 8, 1994

Table 2. Results of Itaconic Acid Experiments expt no. reactant

catalyst

conv res time ( 8 ) (%) COz CO C a b MA IC CC MC CT A0 Effect of Temperature 83 47 4 2 20 18 11 10 1 4 64 90 66 6 2 26 13 10 8 2 5 62 91 67 5 2 26 12 9 6 2 6 56 92 67 6 2 30 10 8 4 4 6 54 9686 4 1 38 4 4 3 1 3 61 9893 4 2 42 2 2 2 1 3 61 991004 1 44 1 1 1 2 3 58 4 46 1 1 1 2 4 99 111 6 55 45 0 0 0 2 6 100 110 4 7 49 0 53 2 7 0 0 0 3.5 9853 0 2.1 59 3 12 10 0 0 97 56 0 0 33 4 9 5 0 0 9663 1 0 3.5 40 3 6 3 0 0 9755 0 0 3.2

20 21 22 23 25 26 27 28 29 55 49 56 58

0.5 MIC 0.5 M IC 0.5 M IC 0.5 M IC 0.5MIC 0.5MIC 0.5MIC 0.5MIC 0.5MIC 0.01MIC O.01MIC O.01MIC 0.01MIC

nil nil nil nil 10mMNaOH 10mMNaOH 10mMNaOH 10mMNaOH 10mMNaOH 10mMNaOH 10mMNaOH 2mMNaOH 2mMNaOH

360 375 385 400 360 365 370 375 400 360 375 360 375

24 32 27 33

0.5MIC 0.5MCC 0.5MIC 0.5MCC

10mMNaOH 10mMNaOH 10mMNaOH 10mMNaOH

350 350 370 370

71 67 58 58

54 81 82 104 102 34

0.05MIC 0.05MIC 0.05MIC 0.05MIC 0.05MIC 0.05MIC

10mMNaOH 10mMNaOH 10mMNaOH 10mMNaOH 10mMNaOH 10mMNaOH

360 360 360 360 360 360

3.5 5.1 9.2 25 27 62

112 31 24 30 71 20 25 35 64 80 73 109 72 110 65 79

0.05 M IC 0.5MIC 0.5MIC 0.5MIC 0.5MIC 0.5 M IC 0.5MIC 0.25MIC 0.05MIC 0.05MIC 0.5MIC 0.2MIC 0.2MIC O.1MIC 0.10MIC O.1MIC 0.05 M IC 0.05 M IC 0.05 M IC 0.05MIC 0.05MIC 0.05MIC O.1MIC 0.05MIC 0.05MIC 0.01MIC O.1MIC O.1MIC O.01MIC

10 mm NaOH 10mMNaCl 10mMNaOH 10mMKOH 10mMH2S04 nil 10mMNaOH 10mMNaOH 5mMNaOH 5mMNaOH 100mMNaOH 40mMNaOH 40mMNaOH 20mMNaOH 20mMNaOH 20mMNaOH 10 mM NaOH 10 mM NaOH 10 mM NaOH 10mMNaOH 10mMNaOH 10mMNaOH 20mMNaOH 10mMNaOH 25mMNaOH 10mMNaOH 10mMNaOH 50mMNaOH 50mMNaOH

350 350 350 350 360 360 360 360 360 360 360 360 360 360 360 360 360 360 360 360 360 360 375 375 375 375 375 375 375

10 70 71 66 60 64 61 61 27 25 25 10 25 10 27 25 10 25 27 25 25 62 21 21 21 20 2.1

111

104 102 78 86 34 74 75 76 77 48 50 51 a

1.1 1.1

Effect of Reactant 2 34 7 7 3 1 9378 0 9372 2 3 33 7 6 5 1 991004 1 44 1 1 1 2 9986 3 8 44 1 1 1 2 Effect of Residence Time 0 0 31 7 19 11 0 93 29 92 56 0 0 50 8 11 6 1 9879 0 0 66 2 2 1 2 1 71 0 0 0 4 10095 4 1 70 0 0 0 4 10090 2 1001003 3 72 0 0 0 0 Effect of pH and Reactant Concentrations 90 46 1 0 46 10 15 9 1 80 42 2 0 20 20 14 14 0 9378 0 2 34 7 7 3 1 9367 3 4 32 7 6 5 1 80 39 11 0 14 20 19 13 0 83 47 4 1 20 18 11 10 1 9686 4 1 38 4 4 3 1 9996 4 2 60 1 1 1 2 98 nda nd nd 74 2 0 0 2 9882 0 1 60 2 2 1 3 1 57 1 0 0 1 99 102 3 93 78 0 0 53 7 10 5 3 1 67 1 0 0 1 99 105 3 93 53 1 0 54 7 11 5 2 99 nd nd nd 75 1 0 0 2 9986 4 1 72 1 0 0 2 1 1 60 6 10 4 2 94 58 1 71 0 0 0 4 10095 4 10090 2 1 70 0 0 0 4 9989 0 1 72 1 1 1 0 9892 5 2 63 2 0 0 2 1001003 3 72 0 0 0 0 99 103 3 1 67 1 0 0 2 1 70 1 0 0 2 9996 3 9998 2 2 69 1 0 0 1 1001000 3 65 0 0 0 0 80 20 0 0 19 20 33 16 0 89 36 0 0 36 11 28 10 0 75 0 0 0 0 24 18 52 0

AA AD PY HIB Cm

sample

2 4 1 2 7 6 4 6 1 3 8 0 3 7 0 3 7 6 5 1 0 0 3 7 8 3 2 1 4 6 6 3 2 0 5 7 2 5 7 0 3 3 0 3 0 0 4 7 8 3 1 0 3 7 1 27 0 25 3 89 11 0 13 5 89 33 0 17 5 83 32 0 16 5 81

10/27/92 N4, N6 10/27/92 N8, N10 10/27/92 N20, N22 10/27/92 N30, N28 12/03/92 Bl-B4 12/03/92 C 1 4 4 12/03/92 Dl-D4 11/17/92 K7-KlO 11/17/92 K13-Kl6 4/7/93 nl, n2 3/22/93 s&slO 4/7/93 m2, m3 4/8/93 83 84

2 3 3 3

3 3 3 3

2 2 3 2

1 4 6 6 1 4 6 8 0 5 7 0 0 4 6 9

12/03/92 Al-A4 12/23/92 B2-B5 12/03/92 Dl-D4 12/23/92 Dl-D4

0

8 9

0 0 0 0 0 0

5 4 7 5 6 9 7 8 9 2 1 0 9 4 1 0 9 6 9 9

4/7/93 sl,s2 5/7/93 83-85 5/7/93 81, s2 8/9/93 81, s2 6/23/93 85, e6 2/4/93 s17, a19

3 4 9 7 0

11

3 8 7

3 1 6 4 3 2 2 3 2 2 3 1 10 3 0 4 2 4 3 3 2 3 6 4 n d 8 0 3 12 8 4 2 0 7 2 4 5 2 0 3 1 4 2 n d 3 1 4 4 2 5 1 2 9 3 0 7 8 0 4 7 4 2 11 0 0 7 0 4 4 2 5 6 5 4 6 2 0 14 0 0 3 0 0 2 0 0 6 0

7 1 2 5 6 3

2 0 1 1 0 1 1 2 3 4 1 2 2

5 9 7 2 7 9 4 6 6 3 6 6 0 8 1 2 7 6 4 6 6 1 5 9 1 7 n d 7 9 1 7 7 9 5 9 5 9 8 8 5 9 0 1 7 n d 3 1 0 9 5 2 8 9 5 2 1095 4 1 0 9 6 4 1 0 9 7 5 9 9 0 6 9 9 3 2 8 8 9 5 9 9 7 3 1 2 9 5 7 11 91 1 0 8 8 2 0 8 7 3 0 9 7

10/12/93 a2-a4 1/26/93 88, sll 12/03/92 Al-A4 12/23/92 Al-A4 4/23/93 ml, m2 10127192N4, N6 12/03/92 Bl-B4 2/4/93 e l l , s12 4120193 e l l , 814 5/4/93 814,815 4/23/93 11, I2 10/12/93 d l l - d l 3 4/23/93 nl, n2 10/12/93 C a - C l O 4120193 816, e17 5/4/93 86, e8 10/12/93 b 5 b 7 8/9/93 el, s2 6/23/93 s5,s6 5/4/93 el, 83 5/17/93 811,812 2/4/93 S17, S19 4/28/93 83, a4 4/28/93 ml, m2 4/28/93 nl, n2 4/28/93 11, I2 3/22/93 s3,s5 3/24/93 817,818 3/24/93 820-23

nd, not determined.

Table 3. Other Results expt no.

reactant

catalyst

40 41 42 440 45afb 46avb 47"

MA0.5 M MS0.1 M MS0.1M 0.1 MIC+O.lMMA 0.1 MIC+O.lMMA O.lMIC+O.lMMA O.lMIC+O.lMMA

10mMNaOH nil 10mMNaOH 50mMNaOH 50mMNaOH 10mMNaOH 10mMNaOH

temp res conv (OC) time(s) ( % ) COz 6 370 58 23 67 4 0 350 0 67 2 350 91 82 375 1 89 34 375 1 375 1 80 17 80 21 375 1

CO C& 2 2 0 0 0 0

0 0 0

0 0 0

0 0

MA IC 77 0 0 0 0 0 146 9 137 10 101 20 118 20

CC MC CT A0 0 0 2 2 0 0 0 0 0 0 0 0 20 7 0 0 24 9 0 0 30 12 0 0 31 14 0 0

AA AD PY HIB 0 0 0 5 0 0 0 0 0 0 0 0 3 0 1 6 0 0 2 6 3 0 1 2 3 0 1 4

Cm 87 97 98 99 95 85 95

sample 2/11/93sl,s2 1/26/93sl,s2 1/26/9385,s6 3/4/93810,89 3/4/93~19,820 3/4/93s14,815 3/4/93s2,83

a Yields are based on reactant itaconic acid, whereas C u is based on both reactants. Conducted at pressure 4000 psi instead of normal 5000 psi.

ranging from 220 to 400 O C . The critical temperature of water is 374 "C;consequently reactions conducted below

this temperature involved hot compressed liquid water as the solvent, whereas above this temperature the water

Ind. Eng. Chem. Res., Vol. 33, No. 8, 1994 1993

was supercritical. The high pressure ensured that over the entire temperature range the density of the water was sufficiently high to sustain liquid-phase, heterolytic reactions and minimize the role of gas-phase, free-radical reactions in product formation. Experiments 108, 106, 103, 105, 3, and 4 in Table 1 illustrate the effect of temperature on the reaction chemistry. At 250 OC and below the decarboxylation of citric acid occurs slowly with high selectivity. In experiments 108 and 103 the combined relative yields (relative yield = absolute yield/conversion) of the three propene dicarboxylic acids exceeded 90% ,while in experiment 105 the combined relative yield was about 85%. These results offer an indication of the reproducibility of our work. The yield of itaconic acid is typically a factor of 3 larger than citraconic acid under these conditions. Above 250 O C parasitic reactions begin to evidence themselves in a declining carbon balance accompaniedby the appearance of unwanted byproducts (acetone and acetic acid). At both 230 and 280 "C the addition of base NaOH increases the reaction rate but not the sum of the relative yields of the propenedicarboxylic acids (which remains at about 86% for both experiments 107 and 6). This s u m falls to about 82 % at higher temperatures (see experiments 6-8) in the presence of base. Since it is not difficult to obtain an almost quantitative conversion of citric acid to the propenedicarboxylic acids (primarily itaconic acid), the focus of this paper shifts to the reaction chemistry of these'acids in hot compressed and supercritical water. The first question of interest is the rate of isomerization of itaconic acid relative to its rate of decarboxylation. A comparison of experiments 24 with 32, and 27 with 33 (under Effect of Reactant in Table 2) shows that the rate of interconversion between itaconic and citraconic acids is very high relative to the rates of all other reactions. Almost identical products are obtained independent of which acid is used as reactant. Consequently, these two acids should be regarded to be a single reactant as far as discussions of reaction specificity are concerned. The second question of interest concerns the acidity of itaconic acid in liquid water at high temperatures. According to Le Chatelier's principle it is expected that increasing temperature will affect the dissociation equilibrium according to the sign of A&i. Thus sulfuric acid (whose dissociation is strongly exothermic) becomes a weak acid in supercritical water (Quist et al., 1965; Narayan and Antal, 1990); whereas tert-butyl alcohol becomes a relatively strong acid in compressed liquid water at 250 OC (Xu and Antal, 1994). A qualitative test of the ability of a putative acid to dissociate in liquid water at temperatures above about 300 "C is to examine the stability of 2-propanol in its presence. At elevated temperatures very low concentrations of Arrhenius and Bronsted acids serve as effective catalysts for the quick dehydration of 2-propanol to propene. In the absence of acid, 2-propanol is stable at these conditions. An experiment involving the dehydration of 1.0 M 2-propanol in the presence of 0.1 M itaconic acid at 360 OC, 34.5 MPa, and 27 s residence time resulted in a 70 % conversion of the alcohol to propene (which was the only significant gaseous product). At 375 O C a similar conversion of 2-propanol can be effected by 0.0005 M sulfuric acid after 1.5 s. Thus itaconic acid is weak relative to sulfuric acid at these conditions; nevertheless this result shows that its dissociation must be considered in any examination of its reaction chemistry in water at elevated temperatures. Another important question concerns the stability of

methacrylic acid in water at elevated temperatures. Experiment 40 in Table 3 shows that methacrylic acid decomposes slowly under relatively severe conditions to form propene, crotonic acid, a-hydroxyisobutyricacid, and acetone. If we suppose that the difference between the carbon dioxide and propeneyields is propene which reacted with remaining methacrylic acid to form an undetected ester, then the carbon balance for the experiment approaches unity. Obviously, great care must be taken in the choice of reaction conditions to avoid the degradation of methacrylic acid into unwanted byproducts which are capable of attacking the parent. Among others, experiments 20-29 in Table 2 show that decarboxylation of itaconic acid is fast above 350 "C. Unfortunately, the appearance of parasitic reaction byproducts (such as acetic acid, pyruvic acid, acetone, and acetaldehyde) and methacrylic acid degradation products (propene) signals a loss of selectivity at these temperatures. To improve the carbon balance and to gain insight into the parasitic reaction chemistry, we engaged in a search for undetected and/or unidentified products. Only two, small, unidentified HPLC peaks were found, which did not absorb strongly in the UV. We prepared methyl esters of product samples from the reactor and authentic standards for all identified, significant products and injected these esters into the GC-MSD. No large, unidentified peaks were detected. One small, unidentified peak had an mlz value of 144, which could be the methyl ester of paraconic acid (see Figure 3). One explanation for the lower carbon balance in these experiments involves possible reactions of methacrylic acid with the itaconic/ citraconic acid reactants and/or their other decomposition products. Experiments 44-47 in Table 3 were designed to test this hypothesis. A comparison of the results of experiment 44with 50 and experiment 47 with 48 indicates complete recovery of the methacrylic acid reactant, combined with additional methacrylic acid formed from the itaconic acid reactant. Evidently methacrylic acid is not attacked by the product "soup" that results from the decomposition of itaconic acid. Anticipating that the pH and buffer strength (reactant concentration) of the medium would affect the reaction chemistry, we sought to improve selectivity by systematically varying these at temperatures where the rate of decarboxylationis relatively high. Listed in Table 2 under Effect of pH and Reactant Concentration are blocks of data organized on the basis of increasing temperature. Within each block data are organized on the basis of increasing pH. Short residence time experiments conducted using the capillary tube reactor are grouped separately. As expected (Mok et al., 19891, experiments 71,20, and 25 show that decreasing pH favors the parasitic decarbonylation pathway that results in acetone formation. To verify that the effect of added NaOH in experiment 25 is due to a change in pH, and not the nature of the cation or the general presence of ions, we executed experiments with KOH and NaCl as additives. Results using KOH (see experiment 30) were effectively identical to those with NaOH, whereas those with NaCl (experiment 31) were quite similar to those with pure water (see experiment 20, which was conducted at a slightly higher temperature). From these results we concluded that the cation exerts little influence on the reaction: pH is a controlling variable. In order to verify the proposed mechanism for the transformation of citric to itaconic acid given in Figure 2, we undertook a labeling experiment. At 250 O C , decomposition of labeled material reproduced well the results of

1994 Ind. Eng. Chem. Res., Vol. 33, No. 8, 1994 doublet, J = 6 Hz

1

HocH doublet, J = 12 Hz

' 3 ~ 0 0 ~

-

Ha

\ Hb,c-y13~00H \

singlet

-

6, labelled with 13C

HzC\COOH

A, no 13c

f

Figure 5. Labelingpattern in itaconic acid formed by decomposition of labeled citric acid.

experiment 105 in Table 1. There was 92 % conversion of citric acid, and the major products were itaconic and citraconic acids in the ratio 2.74 (experiment 105: 55% conversion, ratio 2.83). Other products were detected in trace amounts. The crude products were dried for 2 days. A control experiment verified that itaconic acid survives this treatment unchanged. It was the lH NMR spectrum that proved most informative as to the location of the label in itaconic acid. Were the label to be in the carboxyl group adjacent to the methylene, the methylene hydrogens must show splitting by the adjacent 13C (ca. 6 Hz as reported by Abraham and Loftus (1978)). They appear as a singlet; consequently carboxyl carbon "A" in Figure 5 cannot be labeled. By contrast, both vinyl hydrogens (Ha, Hb), singlets in the unlabeled material, are coupled to 13Cin the labeled itaconic acid (&,-l3c= 6 Hz; &,,-l3C = 12 Hz as in Abraham and Loftus (1978)). In the 13C NMR spectrum of unlabeled itaconic acid, the two signals for the carbonyl carbons appear at 6 (DMSO) 167.97 and 172.50 ppm. Only one of the two carbonyl carbon signals is enhanced in the labeled material (6 167.92). The label is only in the conjugated carboxyl group, "B" (see Figure 5),exactly as predicted by the mechanism. Similar results are obtained at a run at 360 "C. Once again only carboxyl group "B" is labeled. These results are consistent with the role of the heterolytic reaction mechanism displayed in Figure 2: absolutely no scrambling of the label was observed. Unfortunately, our drying procedure does not allow analysis of the other major products. These products, citraconic acid at 250 OC, and methacrylic acid at 360 "C, do not survive. The mechanism for decarboxylation of unsaturated acids given in Figure 2 indicates the key role played by the double bond and the @-carboxylichydrogen in the loss of carbon dioxide. The importance of the double bond is illustrated by the results of experiments 41 and 42 in Table 3, which show that methylsuccinic acid (MS) (the saturated analog of itaconic acid, see Figure 3) does not decarboxylate under the reaction conditions. Experiment 51 in Table 2 provides evidence for the role of the @-carboxylichydrogen in decarboxylation: at high pH only isomerization occurs (no decarboxylation is observed). Moreover, the absence of vinylacetic acid (see Figure 3) as a product in all our experiments indicates that only the @-carboxylgroup in itaconic acid leaves as carbon dioxide. The low yields of crotonic acid are also consistent with our expectations: the rate of decarboxylation of citraconic and mesaconic acids to crotonic acid is slow because these two propenedicarboxylic acids lack a double bond @ to their carboxyl groups. The indicated mechanism also helps us to anticipate how an increase in pH (due to a decrease in the acid to base ratio) will affect the rate of decarboxylation of itaconic acid. The carboxylic acid group which is directly attached (a)to the double bond in itaconic acid is expected to be the more acidic as it is closer to the electronwithdrawing double bond; consequently a negligible amount of the carboxylic acid group @ to the double bond should be ionized as long as the pH of the buffer is lower than the pKaof the second dissociation constant at reaction

temperature and pressure. As the double bond becomes more electron rich due to the loss of a proton from the a-carboxyl, it is more likely to form an intramolecular bond with the hydrogen on the @-carboxyl,which initiates decarboxylation. When the pH of the solution exceeds the pKa of the second dissociation constant, then the rate of decarboxylation will decrease as there are fewer 0-carboxylhydrogens available to initiate decarboxylation. Thus the mechanism displayed in Figure 2 leads us to anticipate a relatively narrow range of reactant itaconic acid to base ratios that favor the desired decarboxylation. Because neither the pKal nor the pKd of itaconic acid in near-critical water is known, it is not possible to speak quantitatively of the pH of a near-critical buffer composed of itaconic acid and base NaOH. The situation is exacerbated by the fact that the pH changes as itaconic decomposes to methacrylic acid, whose PKa is also unknown at reaction conditions. Consequently, at present we must be content with phenomenological observations. Experiments 34,64,65,78,79,102, and 104 (see Table 2) show that high yields (exceeding 70%) of methacrylic acid can be obtained at low reactant concentrations (0.05-0.1 M) with reactant to base ratios between 5:l and 1O:l after about 25 s residence time at 360 "C. Higher concentrations of itaconic acid with the same reactant to base ratio result in lower yields of methacrylic acid (see experiments 72, 73, and 109-111). Similarly, higher reactant concentrations with a higher acid to base ratio also result in lower yields (experiments 35 and 25). An increase in temperature to 375 "C at the favored conditions (experiment 75) evidences a slight decrease in methacrylic acid yield, and variations of reactant concentration and acid to base ratio (experiments 74, 76, and 77) at this temperature do not improve the yield. Similarly, a decrease in temperature to 350 "C at the favored acid concentration and acid to base ratio (experiment 112) results in no improvement of the methacrylic acid yield (compare to experiment 111). Reasoning that secondary reactions could be responsible for destroying the desired methacrylic acid product a t long residence times, we conducted experiments with a capillary tube flow reactor that offers very short residence times (