Chemical Conversion of Wood Residues, Part V

Kinetics of the Acid Catalyzed Conversion of Glucose to 5-Hydroxymethyl-2-Furadehyde and Levulinic Acid1

SAMUEL W. McKIBBINS, JOHN F. HARRIS, JEROME F. SAEMAN Forest Products Laboratory,2 Forest Service U. S. Department of Agriculture and WAYNE K. NEILL University of Wisconsin, Madison, Wisconsin

N THE HYDROLYSIS REACTION that The data given in this report are of importance when I converts hexosans to hexoses, a con­ siderable amount of the monosac­ considering the utilization of carbohydrate residues. In charide reacts to form levulinic acid the industrial hydroIysis of starches and agricultural resi­ and 5-hydroxymethyl -2-furaldehyde (HMF), which appear in the product dues to sugars, the production of large quantities of as undesirable contaminants. This re­ 5-hydroxymethyl-2-furaldehyde and levulinic acid results action is of importance in industrial in low yields of the sugars. However, the potential im­ hydrolysis of starch and in the labora­ tory when analyzing vegetable material portance of these two materials as chemical Intermediates for cellulose content. In both cases, the makes them valuable products themselves. This is formation of levulinic acid and HMF especially true in retation to the chemical utilization of represents a loss in yield. Attempts to utilize the carbohydrate wood. fraction of agricultural residues have frequently been directed toward the kinetics of the decomposition of glu- known for over a century. It has been development of processes for the pro­ cose in aqueous acidic media and its rather well established by several duction of simple sugars. Unfortun­ conversion to HMF and levulinic acid. investigators (1, 2, 3)3 that the ately, during hydrolysis of the cellu­ The production of HMF and levu- mechanism of the reaction consists es­ losic material, ther is usually heavy linic acid by the action of hot mineral sentially of a series of consectuive re- destruction of these monosaccharides, acids on various sugars has been actions which follow this order: resulting in low yields of sugar that is contaminated with large quantities of levulinic acid and HMF. Recently, the possibility of produc­ ing these compounds from wood has been investigated in research at the United States Forest Products Labora­ tory, Madison, Wisconsin, into the

1 This is the fifth in a series of reports on chemical conversion of wood residues. Parts I, II, III, and IV appeared in Sept. 1958, May 1959, Feb. 1960, and July 1961, respectively. This is a contributed paper. FOREST PRODUCTS JOURNAL 17 tion. Mechanical strength tests showed that the ampoules could withstand in­ ternal pressures of approximately 1,300 pounds per square inch without failing, and heat transfer tests showed that the ampoules' contents reached the bath temperature in 25 seconds. An appropriate correction factor was applied to all the results to compen­ sate for this short warm-up period. Ampoules were charged with precise amounts of solution from a hypoder­ mic syringe pipe, were chilled to re­ duce the vapor pressure, and then a vacuum applied to remove oxygen and reduce the pressure inside the tube. Fig. 2.— Activity coefficients for glucose While the vacuume was still being a Fig. 1.— Shown is glucose disappearance data and 5-hydroxymethyl-2-furaldegyde disappear­ plied, the tubes were sealed off and the at 180° C. and 0.2 N. sulfuric acid. ance are shown. ampoule was completed. This loading technique showed variations in the volume charged of less than 0.2 per­ cent and was rapid enough to allow They suggest that the reaction pro­ that the reaction mechanism was preparation of 200 ampoules per day. ceeds through several intermediate complicated. Ampoules were reacted in a thermo­ compounds and that one mole each of statically controlled bath of hydro­ HMF levulinic acid, and formic acid is Experimental and Analytical genated cottonseed oil for specific time produced per mole of initial glucose. Techniques intervals and were then quenched in a The overall kinetics of this series of second bath. Transfer of ampoules reactions has not been developed in This investigation was carried out in from one bath to the other was per­ the past, but several of the individual two main studies. In the first, the dis­ formed behind safety shields with the steps have been investigated. The dis­ appearance of gluclose and the simul­ aid of rotating brackets with control appearance of glucose in dilute solu­ taneous formation of HMF and levu­ rods. The heating bath was completely tions of mineral acid has been fol­ linic acid was measured. Measure­ encased to contain any explosions re­ lowed (4, 1, 5) at temperatures below ments were also taken on the forma­ sulting from the occasional failure of 200°C., and the disappearance of tion of the total organic acids and the an ampoule. The temperature of the HMF in dilute mineral acids has also insoluble solid materials. In the sec­ bath was controlled within 0.1 C.° at been investigated (6, 1, 7). It has ond, the disappearance of HMF and the lower temperatures and 0.2 C.° at further been shown (1, 8) that levu­ the simultaneous formation of levu­ the higher temperatures. linic acid displays no marked tendency linic acid was determined. The range The contents of the ampoules were to decompose at these reaction of variables for these studies included obtained for analysis by breaking open conditions. temperatures of 140°C. to 250°C., both ends and transferrin the reacted Most of the previous work in this catalyst acid concentrations of 0.025 solututions to vials. The analysis for glu­ field, however, was concerned primar­ N. to 0.8 N. sulfuric acid, and sev­ cose was that described by Shaffer and ily with the determination of yields of eral concentrations of initial glucose Somogyi (9) and depends on the re­ HMF and levulinic acide produced and HMF. ducing power of glucose. The reacted from various sugars and the reaction The experimental data were ob­ solutions were analyzed directly for conditions necessary to obtain these tained in batch reactors which were residual glucose since there were no yields. The variations of the yields heated in a constant temperature bath reaction products present which would with time, as are necessary in a kinetics for various time intervals. Thereactors interfere. The reducing power of study. were in general not included. were glass ampoules made from 5­ HMF was found to be 5 percent of In all cases, the yields were less than millimeter (Pyrex) glass tubing which that of glucose, and it was present in the theoretical, and it was concluded had a capacity of 1.0 milliter of solu- too small quantities to influence the measurements. Due to the presence of materials which would interfere in subsequent analyses, the HMF was first chromato­ Table 1.—PLATEAU VALUES OF LEVULINIC ACID YIELDS graphically separated on paper from

* 1 the reaction mixture. The solvent sys­ FROM GLUCOSE, (CL )p, (moles/mole) tem employed was the organic layer Sulfuric acid concentration, CA , gram-equivalent per liter resulting from an equal volume mix­ Temperature 0.05 0.025 ture of five molar formic acid and °C. 0.8 0.4 0.2 0.1 pentanol. A standardization of the pro­ cedure with known quantities of HMF showed that its recovery from the pa­ per was quantitative at 92 percent with a standard deviation for the experi­

2 Maintained at Madison, Wisconsin, in 1C = 0.556 gram-molecules per liter except as noted. cooperation with the University of Wis­ G o 2C = 0.278 gram-molecules per liter. consin. G o 3Numbers in parentheses refer to Litera­ 3C = 1.112 gram-molecules per liter. Go ture Cited at the end of this report. 18 JANUARY, 1962 Fig. 3.— Effect of temperature on the yield of 5-hydroxymethyl-2­ Fig. 4.— Shown are maximum yields of 5-hydroxymethyl-2­ furaldehyde from glucose is shown (CG = 0.556 gram-molecules per liter). o furaldehydeproducedfromglucose.

mental points of 1.59-percent. The in­ ness at room temperature since the am­ termined. A sample plot is shown in dicated 8 percent loss is probably due monium ions would interfere with the Figure 1. The rate constants were cor­ to its volatility since tests showed that subsequent analysis. The method em­ related as prescribed by the Arrhenius less than 1 percent of the applied ployed to measure the sodium levulin­ equation from which an activation en­ HMF was located outside the band ate concentration of the separated solu­ ergy of 32,690 gram-calories per eluted for analysis. The solutions of tion was suggested by Ploetz and Bar­ gram-molecule was obtained. separated HMF were then analyzed for tels (10) and depends on the acetyl To incorporate the proportional de­ optical absorbance at 284 mµ. with a group, which undergoes the iodoform pendency of the rate constants on the Beckman model DU spectrophotome­ reaction in the presence of alkaline catalyst acid concentration, an activity ter to determine the concentration. iodine solutions. The stoichiometry of term was included in the coefficient of An investigation of the optical prop­ this reaction has not been satisfactorily the Arrhenius equation. The standard erties of the known and suspected re­ developed, and hence a calibration was state for the activity coefficient. aG, action products and their behavior made with pure levulinic acid under was arbitrarily defined as unity at with the solvent system mentioned, conditions which were stringently ob­ 180°C., a catalyst concentration of showed that they either would not served in subsequent work. This cali­ 0.8 N., and an initial glucose concen­ separate with HMF or they would not bration indicated that 7.18 equivalents tration of 0.556 gram-molecule per absorb significantly at 2M mµ. Also, of iodine are consumed per equivalent liter. The values of the activity coef­ a study of the optical spectrum of of levulinic acid The standard devia­ ficients at other conditions were then HMF separated from reaction mixtures tion of the experimental points in this calculated using the experimental val­ did not indicate the presence of con­ calibration was 1.74 percent. ues of k1. Over the range of tempera­ taminants. Due to a tendency of the The analysis for the total organic tures and acid concentrations investi­ dilute HMF solutions to degrade to 5­ acids present consisted of titrating the gated, aG was only a function of hydroxymethyl-2-furoic acid at room reacted samples with a standardized catalyst concentration and showed only temperatures and when exposed to base and subtracting the acidity of the minor, random variations with temper­ light, it was necessary to store them catalyst acid, and the amount of solids ature. This relationship is shown in in a dark refrigerator until they could formed was found by filtering, drying Figure 2. It should be noted that the be analyzed. and weighing the solids from a reacted activity coefficient, although it re­ The levulinic acid analysis was also ampoule. sembles that for sulfuric acid, is de­ preceeded by a paper chromatographic fined for the solution as a whole, separation. In this separation, the sol­ Glucose Disappearance hence, the subscript G. vent system consisted of ethanol (95 It was found that the glucose dis­ Most of the experimental work was percent), aqueous ammonia (29 per­ appearance follows a first-order carried out at an initial glucose con­ cent), and water in a volume ratio of mechanism: centration of 0.556 gram-molecule 100 :5 :5, respectively. Due to its per liter with a few additional runs at volatility with this solvent, the levu­ 0.278 and 1.116 gram-molecule linic acid was chromatographed as its (1). per liter to determine the effect of ammonium salt which separated into a varying the initial glucose concentra­ narrow, distinct band. After elution of tion. At all concentrations a first-order * this band, it was in turn converted to Experimental values of ln C G plotted disappearance occurred; that is, the * the sodium salt and the ammonium against time resulted in straight lines plots of ln C G versus time were all ions removed by evaporating to dry­ from which the values of k1 were de­ linear. However, as seen in Figure 1, FOREST PRODUCTS JOURNAL 19 Fig. 6.—Shown is the effect of sulfuric acid concentration on the

yield of levulinic acid produced from glucose (CG = 0.556 gram- Fig. 5.—Correlating function A is shown. molecules per liter). o the slopes of these lines and, conse­ time. This term was used because it vary from one-third to one and one- quently, the values of k1 were not the allowed all the data to be shown on third glucose half-lives. These yields same for various concentrations. It is one graph. It is related to the glucose may also be converted to moles of held, nevertheless, that the reaction is disappearance half-life, t1/2, thus: HMF formed per mole of glucose re­ a true first-order and the discrepancy is acted with the aid of equation (1). attributed to the manner in which the On this basis, the maximum HMF solutions were made up. This appar­ yields vary from 0.18 to 0.27 moles/ ent discrepancy is discussed in a later In Figure 4 are plotted the maxi­ mole. section. It was consequently necessary mum yields of HMF where it is shown If one neglects the presence of in­ to include the initial glucose concentra­ that the yields are increased with in­ termediates and assumes first-order re­ tion as a parameter in the activity creased temperature and decreased ini­ actions exclusively, the following sim­ coefficient plot. tial glucose concentration, but are in­ plified mechanism can be written: The corrected Arrhenius equation dependent of catalyst concentration ex­ has this form: cept at higher temperatures where they are increased with decreasing amounts Glucose HMF of arid. A possible explanation of this effect is given later. The times at Assuming constant volume and the which these maxima occur may be boundary condition that the HMF

calculated using equation (4) below concentration, CH, equals zero at time (2). and will be seen to shift to longer re­ equals aero, the dependence of CH It may also be noted that the glu­ action times as the yields increase and with time is: cose disappearance curves have inter­ cepts at time qual to zero that de­ (3) crease as the concentration increases. This is probably due to the reversion of the glucose to polysaccharides which has been shown to be more pro­ nounced at higher concentrations. This Table 2.—MAXIMUM YIELDS OF TOTAL ORGANIC ACIDS FROM does not affect the slopes and, conse­ GLUCOSE (C* ) , 1 (moles/mole) quently, the rate constants, however. 0A m

Sulfuric acid concentration, CA, gram-equivalent per liter HMF Formation from Glucose Temperature The formation of HMF from glu­ °C. 0.8 0.4 0.2 0.1 0.05 cose shows a growth and decay with time as typified by Figure 3. The data are reported as moles present at any * time per mole of initialglucose, C H, and represent the fraction of the theo­ 1C = 0.556 gram-molecules per liter except as noted. retical yield. The time function is a G o 2 dimensionless time, k1t, obtained by C = 0.278 gram-molecules per liter. G o 3 multiplying the glucose disappearance CG = 1.112 gram-molecules per liter. rete constant by the isothermal reaction o 20 JANUARY, 1962 Fig. 7.—Shown is the effect of sulfuric acid concenstration on the Fig. 8.— Solids formation from glucose is shown. yield of total organic acids produced from glucose (CG = 0.556 gram-molecules per liter, 200° C.).

where a = k2 /k1. The time for maxi­ At catalyst concentrations less than 0.1 and is independent of catalyst concen- * mum formation of C , (k1t) m, obtained N., however, this relationship no tration except at low concentrations H longer holds. where it decreases with decreased amounts of catalyst. Values for from equation (3) is: The correlating equation was fur- ther modified by the inclusion of a constant coefficient to yield this final (C * ) are given in Table 1. (4), form: L p The simple two-step mechanism con­ and the value of maximum C * , (C * ) , H H m (6). sidered earlier can be extended to in­ is found by substituting equation (4) clude levulinic acid where only first- into quation (3). order reactions are considered: Due to the presence of intermedi­ This equation is shown plotted in Fig­ ates and side reactions, the experi­ ure 3 where it can be seen that it Glucose HMF levulinicacid. mental data were not described pre­ accurately describes the experimental cisely by the above equations. These data. data did display the same general It was found in this investigation that trends predicted by this simple two- Levulinic Acid Formation the levulinic acid displayed no tend- step mechanism, however, and were From Glucose ency to decompose and, therefore, no therefore fitted with empirical equa­ The formation of levulinic acid degradation is included above. Assum­ tions of the same general form using from glucose displays a plateau value, ing constant volumes and the bound- a correlating function A in place of as typified in Figure 6, due to the sta- ary condition that CH and CL are zero the term a. The values of A were ob­ bility of levulinic acid to these reac- when time is zero, the dependence of tained from equation (4), using the tion conditions. The data are reported CL with time is: experimental values of (k1 t)m and, be­ cause of their implied relationship to the rate constants, they correlated nicely with the reciprocal of the ab­ (7). solute temperature as shown in Fig­ ure 5. as moles of levulinic acid per mole of As with the HMF data, it was At catalyst concentrations of 0.1 N. found that the reactions involved were and greater, the relationship between * initial glucose, C , and since the yield too complicated to lend themselves to A and temperature may be represented L an exact discussion of their mechan­ by this equation: * reaches its plateau values, (CL ) p , at ap- isms, and consequently an empirical proximately six glucose half-lives, the approach was used. Equation (7) was plateau value is also the moles of levu- modified by substituting the correlating A= d e (5), linic acid per mole of glucose reacted. function A discussed earlier for a and Mechanisms proposed by previous in- multiplying the right side by the pla­ where: d = 1.64 10–5 when: C × G vestigators suggest that we mole is o * * = 0.278 gram = molecules per liter, formed per mole of glucose. From the teau value of C L , (CL ) p. This latter size of the yields it can be inferred modification was necessary since the d = 2.03 10–5 when: C that there are complicating side reac- × G * o tions. The time in this case is also values of C did not approach unity at = 1.112 gram = molecules per liter, L * large values of k1 t as demanded by –5 equation (7), but reached lower val­ d = 1.76 10 when: C (k1 t). It was found found (C L ) p in- × G o creases with decreased temperature and ues which depended on the reaction = 0.556 gram = molecules per liter, decreased initial glucose concentration conditions. FOREST PRODUCTS JOURNAL 21 Table 3.—PLATEAU VALUES OF LEVULlNlC ACID FROM 5-HYDROXYMETHYL-2-FURALDEHYDE (moles/ mole) (CH = 0.0805 gram = molecules per liter) o Sulfuric acid concentration, CA, gram-equivalent per liter Temperature °C. 0.8 & 0.4 .02 .01 0.05 0.025 (8).

* With the values of (C L ) p from Table 1, equation (8) is shown plotted with the experimental data in Figure 6. A good fit is obtained not and 0.2 N.—0.4 N. sulfuric acid). These data were not comprehensive only at the plateau, as is expected, but These data are shown plotted in Fig- enough, however, to determine if the also between time zero and the pla­ ure 8. disappearance were truly first-order or teau. In extending the values of pseudo first-order and the solutions were too dilute to be expected to show ( C * ) to other reaction conditions, it 5-Hydroxymethyl-2-Furaldehyde Dis- L p appearance and Simultaneous Ap- the variation with initial concentration * pearance of Levulinic Acid which was found in the glucose disap­ should be noted that ln (C L ) p plotted pearance study. versus (1/T) using initial glucose con­ The HMF disappearance followed a The formation of levulinic acid centration as the parameter results in a first-order mechanism; that is, plots of linear plot which is independent of from HMF displays a plateau value as ln (C H / CH ) versus time yielded a function of time. These plateau catalyst concentration for values of o yields, shown in Table 3, display the CA greater than 0.1 N. straight lines, where CH is the initial same general trends with temperature o and catalyst concentration indicated for concentration of HMF. The HMF dis- Organic Acid and Solids Formation levulinic acid previously. The mole From Glucose appearance rate constants, k2,obtained fraction yields, however, are approxi­ from the slopes of these lines were mately twice as large in this study, but The data on the formation of or­ correlated with temperature by means still less than the theoretical. The effect ganic acids from glucose arc reported of the Arrhenius equation and an ac- of varying the initial HMF concentra­ as moles of organic acids per mole of tivation energy of 23.110 gram- tion was not fully determined nor was calories per gram-molecule was a correlation justified in this study. * determined. initial glucose, C0 A. TO do this, it was assumed that all the acids present were To incorporate the dependency of Discussion single dissociation acids since it was the rate constant on catalyst concentra­ believed that formic and levulinic acids tion, an activity coefficient, aH, was The experimental data are corre­ were the main acid materials present. included in the coefficient of the Ar- lated with empirical equations based The yield curves as a function of time rhenius equation which was arbitrarily on a simple first-order consecutive re­ display a maximum with only a slight defined as unity at 180°C. and CA = action mechanism. Because this mech­ tendency to decay as typified by Fig 0.8 N. From the experimental value anism itself does not adequately de­ 7. This is probably predominantly due of k2 at these conditions, the numeri- scribe the true process, it is necessary to the disappearance of the formic acid cal form of the equation was to incorporate correlating functions to which is formed. Earlier mechanisms determined: account for the deviations from this suggested that one mole each of formic and levulinic acids were formed per (9). mole of initial glucose. It was found that the yields were considerably be­ low two, as predicted by this assump­ The values of aH were then calculated assumption. When the curves pre­ tion, and that they were also greater at the other conditions from the ex­ dicted by the equations are compared than twice as large as the yield of perimental values of k2. Over the to the experimental data, however, it levulinic acid. The values of the max­ range of variables investigated, aH was must be concluded that the trends are ima are listed in Table 2 where it will a function only of catalyst concentra­ accurately described. be noted that the maximum yields are tion and showed only minor, random Although it would normally be de­ increased with decreased temperature, variations with temperature. This rela­ sirable to use equations based on the decreased initial glucose concentra­ tion between aH and CA is shown in true mechanism, the empirical ap­ tion, and increased catalyst concentra­ Figure 2. As with the function, aG, it proach was necessary in this study be­ tion, Since the maxima do not occur slould be noted the aH is defined for until six glucose half-lives, these val­ the entire solution, hence the subscript cause the reactions are very complex ues are also essentially moles formed H. and there was not sufficient informa­ per mole of glucose reacted. Most of the experimental work in tion available to adequately determine the true reaction phenomena. More­ Approximately 25 percent of the this study was done at an initial HMF over, even if the actual mechanism initial glucose is converted to solid concentration of 0.0805 gram = mole­ materials during the course of the re­ cules per liter, since this approximated were known, the equations necessary action. The formation curves when the concentration of the HMF in the to describe a process of this complexity would, in themselves, be so compli­ plotted versus (k1 t) are a function solutions in the glucose disappearance only of the initial glucose concentra­ study. When the concentration was cated as to seriously hinder their appli­ tion and are independent of tempera­ varied from 0.061 to 0.139 gram = cation in any subsequent design work. ture and catalyst concentration over the molecules per liter, only minor varia­ For example, the differential equations range investigated (180°C.–220°C. tions were noted in the values of k2. which describe even the relatively sim-

22 JANUARY, 1962 ple mechanism given below are non­ products, since the HMF disappear­ linear. ance study indicated a first-order decay The effect of the initial glucose con­ for HMF itself. A possible mechanism centration on the rate of glucose dis­ is: k2 appearance is attributed to the manner k k3 in which the solutions were made up. Glucose 1 I HMF levulinic acid, The solutions were prepared by weigh­ ing into a volumetric flask the desired amount of glucose, adding the desired k amount of 1.00 N. sulfuric acid, and 4 diluting with water. This procedure solids neglects the change in density of the and identified, their presence was dem­ solutions due to the glucose concentra­ where I is an intermediate; k1, k2, and onstrated. The chromatographic sep­ ion and, hence, the acid to water ratio k3 are first-order constants; and k4 is a is higher in the solutions of higher higher order rate constant. A mech­ arations employed in the HMF and levulinic acid analyses both showed glucose concentration, although the anism of this type would also be re­ normality remains the same. The sponsible for the decrease in levulinic the presence of several unaccounted change in rate may be explained by a acid yields as the initial glucose con­ for compounds in appreciable concentrations. change in activity of the acid as the centration is increased and for the gen­ acid to water ratio varies. eral lowering of the yields from the theoretical value. NOTATION: This is further verified by the work The solids have been shown as the A = correlating function (dimension­ of Kirby (11) in which the glucose probable result of the interaction be­ less) disappearance rate constants did not tween I and HMF since the data on C = concentration (gram = molecules vary over an eight-fold change in ini­ solids formation indicate that its for­ per liter) tial glucose concentration. His solu­ mation occurs by a higher order C* = dimensionless concentration tions were made up using different (C/C G ) mechanism. Most of the data collected o weights of glucose and sulfuric acid on solids formation are shown in Fig­ CA = catalyst concentration (gram = of the same normality. This resulted ure 8. The sulfuric acid concentrations equivalent per liter) in solutions of varying final acid nor­ are all 0.2 N. or 0.4 N. and show no 5-hydroxymethyl-2-furaldehyde malities, but in a constant ratio of acid HMF = influence on the amount of solids pro­ k1 first-order glucose disappearance = –1 to water. The solutions were made up duced. In several experiments at lower rate constant (minutes ) in this work as described, however, be­ acid concentrations, however, the k2 = first-order 5-hydroxymethyl-2-fur­ cause it was felt at the time that main­ aldehyde disappearance rate con­ yields of solids were greater. It seems –1 taining the normality constant was probable, then, that the formation of stant (minutes ) fundamentally more desirable. the solids is responsible for the de­ t = time (minutes) It was also shown that the glucose crease in levulinic acid yields in a a = solution activity coefficient disappearance was approximately pro­ mechanism similar to that shown portional to the catalyst acid concen­ above, and that this side reaction be­ SUBSCRIPTS: tration. Figure 2 shows that at con­ comes significantly more important at A = catalyst acid (sulfuric acid) centrations greater than 0.1 N. they decreasing catalyst concentrations. G glucose were almost directly proportional, but = The results of the second study in­ Go = glucose present initially at lower values they were not. In dicate that the HMF disappearance H = 5-hydroxymethyl-2-furaldehyde addition, yields of HMF and levulinic 5-hydroxymethyl-2-furaldehyde also appears to follow a first-order H o = acid also show a similar dependence. mechanism, even when the reaction present initially This may be explained in part by con­ was carried to 85 percent of comple­ L = levulinic acid sidering each of the rate constants in­ tion. The yields of levulinic acid in m = maximum volved as composed of three parts: an this study were almost twice as high OA = organic acids acid catalyzed constant, an uncatalyzed as when obtained from glucose, again p = plateau constant, and a base catalyzed constant. pointing out the presence of side re­ S = insoluble solid materials At the higher acid concentrations, any actions in the chain between glucose variation in concentration may not and HMF. The yields were still less Literature Cited cause a large enough change in the pH than the theoretical, however, indicat­ or in the ratio of the basic constant 1. Singh, B., G. R. Dean, and S. M. Can­ ing that side reactions or equilibria tor. 1948. J. Am. Chem. Soc. 70: 517. to the over-all constant to be notice­ also exist between HMF and levulinic 2. Wiggins, L. F. 1948. The preparation able, but at the lower concentrations, acid. These plateau values also showed of levulinic acid from sugar. The Sugar the changes in pH are large enough the trend to become more strongly in­ Research Foundation, Inc., England. to cause this ratio to become signifi­ 3. Wolfrom, M. L., R. D. Schuetz, and fluenced by the catalyst concentration L. F. Cavalieri. 1948. J. Am. Chem. cantly larger. Since different rate con­ at low concentrations and higher Soc. 70: 514. stants are involved in the production temperatures. 4. Saeman, J. F. 1945. Ind. Eng. Chem. of each compound, however, it would The above mechanism is obviously 37: 43. not be expected that the variations in 5. Zheltukhin, D. V. 1953. Zhur. Priklad. an oversimplification of the true reac­ Khim. 26: 882. their yields would show the same tion path, but does help to explain 6. Huzimura, T. 1951. J. Chem. Soc. dependency. some of the observed phenomena. Japan, Ind. Chem. Sect. 54: 271. It was also found that the yields of There undoubtedly are other interme­ 7. Tuenissen, H. P. 1931. Rec. trav. HMF increase as the initial glucose chim. 50: 1. diates in the over-all reaction as well 8 Dunaway, J. W. 1950. M.S. thesis. concentration is deceased. This would as other side reactions. Although none University of Wisconsin. indicate that HMF disappears by a of these intermediates were isolated 9. Shaffer, P. A., and M. Somogyi. 1933. higher order reaction. It appears most J. Biol. Chem. 100: 695. likely that it reacts with one of its 10. Ploetz, T., and H. Bartels. 1941. Ber. The authors are indebted to the Kim­ 74B: 1456. precursors in this reaction rather than berly-Clark Foundation and National Sci­ 11. Kirby, A. M. 1948. Thesis. University with itself or with one of its reaction ence Foundation for financial assistance. of Wisconsin. FOREST PRODUCTS JOURNAL 23