Effect of Cosolvent-Methanol Blends on Water Bottoms Growth in

Phase separation and water bottoms growth in storage tanks containing alcohoVgasoline blends ... gasoline, notably methanol with cosolvents, water bot...
0 downloads 0 Views 832KB Size
Ind. Eng. Chem. Prod. Res. Dev. 1985, 2 4 , 623-630

VLOA = glucose anhydride loss, grams per 100 g of original oven-dry wood WWA = grams of mannose anhydride in the solid residue from 100 g of original oven-dry wood WCA = grams of mannose anhydride in 100 g of original oven-dry wood WZA = grams of mannose anhydride put into solution per 100 g of original oven-dry wood WL = grams of free mannose found in the prehydrolysate solution per 100 g of original oven-dry wood WLA = grams of mannose anhydride found in the prehydrolysate solution per 100 g of original oven-dry wood WK = grams of mannose found in the secondary hydrolyzed prehydrolysate solution per 100 g of original oven-dry wood WT = total grams of mannose ‘n all forms in the prehydrolysate solution per 100 g f original oven-dry wood WTA = total grams of mannose anhydride in all forms in the prehydrolysate solution per 100 g of original oven-dry wood WO = grams of mannose in oligomers in the prehydrolysate solution per 100 g of original oven-dry wood WOA = grams of mannose anhydride in oligomers in the prehydrolysate solution per 100 g of original oven-dry wood WLOA = mannose anhydride loss, grams per 100 g of original oven-dry wood FZ = grams of lignin in the solid residue from 100 g of original oven-dry wood UAA = grams of uronic anhydride in the solid residue from 100 g of original oven-dry wood UK = grams of unknown material in the solid residue from 100 g of original oven-dry wood UKL = grams of unknown material in the prehydrolysate solution from 100 g of original oven-dry wood Registry No. H2S04,7664-93-9;D-Xylose, 58-86-6;D-glucose,

d

823

50-99-7; D-mannose, 3458-28-4; xylan, 9014-63-5; D-glUCan, 9012-72-0; D-“man, 9036-88-8;hemicellulose, 9034-32-6. Literature Cited Arhippainen, 8.; Nevalainen, P.; Marttula, T.; Biom, U-A.; Hanninen, E.; Nikula, M. “Development of the Aiva Prehydrolysis Process”; in “Proceedings of the TAPPI Pulping Conference”, Technical Association of the Pulp and Paper Industry: Atlanta, 1981; pp 399-416. Baker, A. J.; Krcmar, G. F. “Kinetics of the Prehydroiysis of Wood”; unpublished report of the USDA Forest Products Laboratory, 1956, 16 pp. Harris, J. F.; Scott, R. W.; Springer, E. L.; Wegner, T. H. “Factors Influencing Dilute Sulfuric Acid Prehydroiysis of Southern red oak”; in ”Progress in Biomass Conversion”, Academic Press: New York, 1984; Vol 4. Lee, Y. Y.; Lin, C. M.; Johnson, T.; Chambers, R. P. Biotechnol. Bioeng. Symp. 1979, 8 , 75. Liegel, E. A.; Simson, C. R.; Schuite, E. E., Eds. “Procedures for Soii Testing, Plant Analysis and Feed and Forage Analysis”; Dept. of Soii Sciences, University of Wisconsin Extension: Madison, WI, 1980; “Soil Fertility Series No. 6”,p 24. Limbaugh, M. L.; Chambers, R. P.; Kaiiianpur, C.; Lee, Y. Y. Sun 2, Proc. Int. Sol. Energy SOC. Sliver Jubilee Congr., 7979 1979, 79, 93-98. Root, D.F.; Saeman, J. F.; Harris, J. F.; Neiii, W. K. For. Prod. J. 1959, 9(5), 158. Rydhoim, S. A. “Pulping Processes”, Interscience: New York, 1965; Chapter 9, p 649. Scott, R. W. Anal. Chem. 1978, 48, 1919. Scott, R. W. Anal. Chem. 1979, 57, 936. Scott, R. W.; Wegner, T. H.; Harris, J. F. J . Wood Chem. Technol. 1983, 3(3), 245. Springer, E. L. Tappi 1966, 49(3), 102. Springer, E. L. Cellul. Chem. Technol. 1983, 77(5), 525. Springer, E. L.; Harris, J. F. Ind. Eng. Chem. Prod. Res. Dev. 1985, 24, 485.

Springer, E. L.; Libkie, K. A. Tappi 1980 63(7), 119. Springer, E. L.; Zoch, L. L., Jr. Tappi, 1988, 57(5), 214. Springer, E. L.; Harris, J. F.; Neill, W. K. Tappi 1963, 46(9), 551.

Received for review May 13, 1985 Accepted August 16, 1985

Effect of Cosolvent-Methanol Blends on Water Bottoms Growth in Gasoline Storage Tanks Frank W. Melpolder ARC0 Chemical Company, Division of Atlantic Richfieid Company, Newtown Square, Pennsylvania 19073

Phase separation and water bottoms growth in storage tanks containing alcohoVgasoline blends represent a critical problem in service station operation. This phase separation on exposure to water is lessened by utilizing optimum blends of methanol with higher alcohols available from natural products, petroleum, and natural gas sources. Ethyl, isopropyl, n-butyl, isobutyl, and tert-butyl alcohols are examined as possible cosolvents with methyl alcohol. The effectiveness of the various cosolvents is compared in terms of maximum water bottoms and number of refills to miscibility. Regression equations were incorporated into a computer program that simulates the successive extractions in a gasoline storage tank. Computer examples demonstrated that the cosolvent ethyl alcohol would produce the highest water bottoms and the C, alcohols the least, with fee-butyl alcohol being the optimum choice.

Introduction

Water accumulation at the bottom of gasoline storage tanks has been a troublesome problem to the facility operator since the beginning of the gasoline industry. At present, with the addition of oxygenated supplements to gasoline, notably methanol with cosolvents, water bottoms growth in tanks at service stations, pipeline terminals, and bulk distribution centers has taken on added significance. In order to prevent excessive growth and accidental delivery of water bottoms to the user, proper operating guidelines must be established based on the solubility characteristics of the major components in the fuel. 0196-4321/85/1224-0623$01.50/0

Keller (1978) has reviewed in detail for the period prior to 1979 the use and some of the physical properties of alcohol fuels. Cox (1979) examined the phase relationships of the gasoline-methanol-water system. Roehm (1977) determined the beneficial effect of adding isobutyl alcohol to methyl alcohol-gasoline blends. Haq (1981) measured the water tolerance of 20% alcohol-gasoline blends at different temperatures and aromatic contents. Ruiz-Bevia and Prats-Rico (1983) studied another type of quaternary system and attempted to establish a geometrical method for prediction of the equilibria based on ternary data. However, the correspondence between predicted and ex0 1985 American Chemical Society

624

Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985

Table I. Physical Properties of t h e Base Fuel Reid vapor pressure (dry), psi vapor/liquid ratio at 20 O F D86 distillation, O F vol % evaporated, IBP 20 50 80 EP specific gravity, 60/60 water (Karl Fischer), wt % octane no., research road sulfur, wt % hydrocarbon types, vol %, paraffins olefins naphthenes aromatics heat of combustion, Btu/lb

Cosolvent

11.5 122 82 134 211 310 418 0.7314 0.005 91.5 82.9 0.130 57.9 11.7 4.5 25.9 20065

perimental was not satisfactory. Experimental Section The work described here is an investigation of the solubilities and partitioning of methanol, cosolvent, water, and gasoline mixed at a range of ratios. The cosolvents were ethyl, isopropyl, n-butyl, isobutyl, and tert-butyl alcohols. Multiple extractions of water bottoms with fuel were performed to simulate field experiences in storage tanks. Correlations of the data were made with a computer in preparation for the development of a computer model that would predict water bottoms growth and extinction. Materials. A nonleaded gasoline typical of marketed motor fuel was selected as the base for the solubility, partitioning, and extraction studies. Detailed analytical inspections that describe the gasoline are given in Table I. Methanol and all the cosolvents for the program were at least 99% purity grade and were used as received from commercial sources without further purification. Procedures. The objective of the experimental work was to supply sufficient data to produce meaningful correlations and to verify computer predictions of water bottoms growth. Concentration maxima were chosen for methanol (11vol %), cosolvent (13 vol %), and water 1.5 (vol %) in preparation of the gasoline blends. The temperature for all the tests was fixed at 22 f 1"C. Although it would be desirable to examine other temperatures and other base gasolines with different aromatic contents, the extra effort and complexity to the program could not be justified. Partition experiments were made with 100-mL volumes of the mixtures contained in graduated centrifuge tubes that were sealed with serum caps. The charge mixture was equilibrated by agitation and allowed to settle or centrifuged until the phases were clear. The volume of the water bottoms phase was recorded, and analytical samples of both phases were withdrawn with syringes. The organic components were analyzed by gas chromatography, and water was analyzed by the Karl Fischer method. Water saturation tests (water tolerances) were made by titration of water from a microsyringe into a septum bottle containing the mixture. Delivery of each 0.02-0.2 mL of water to the system was followed by shaking or with ultrasonics to hasten solubility. The water saturation point was judged to be reached when fine droplets were observed that could not be dissolved. Multiple extractions that were designed to simulate phase behavior in storage tanks were performed in tapered 2-L graduated vessels. Provision was made to syphon off 75-85 vol % of the gasoline phase. The feed mixture was agitated and allowed to settle 1-16 h until clear. Any water hang-up on the side walls was dislodged by swirling and

Figure 1. Effect of cosolvent-methanol ratio on envelope shape. Conceptual view of surface inside tetrahedron shell representing solubility limits for gasoline, cosolvent, methanol, and water.

Cosolvenls Ethanol

isopropanol

"

n .E I a n0I

IsoButanol I.Bulano1

Figure 2. Equilibrium-phase compositions and tie lines for systems ' cosolvent = 5, methanol = 5, water = 1, and gasoline with total vol % = 89.

gentle knocking of the vessel. The bottom layer volume was recorded, and a portion of the upper phase was syphoned off and replaced with an equal volume of fresh fuel. This procedure was repeated until the water bottoms was completely absorbed. Experimental Results. A supplementary table of the experimental data listing the composition, partition coefficients, Ki, and observed water bottoms for gasoline blends of methanol and each cosolvent is available. (See paragraph at the end of the paper regarding supplementary material.) Compositional data for quaternary systems require a 3-dimensional representation. For a typical water-soluble cosolvent, the solubility envelope for methanol-cosolvent-gasoline-water is illustrated in Figure 1 as a curved surface inside a tetrahedron with the pure components designated at the apexes. The broad space above the surface representing the miscible region lies near the methanol and cosolvent apexes, but the soluble region diminishes rapidly as the concentrations of either gasoline or water become large. The open space below the surface, the immiscible two-phase region, shows tie-line locations that connect the gasoline phase composition at one side of the surface with its corresponding water-phase composition at the opposing surface. An attempt is made in Figure 2 to show tie lines in a triangular plot by combining water and methanol at one apex. These lines indicate the

Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985 625 10

10

9

9 8

8 I

c 7

5 7 W

26% Aromatics in Fuel

>

26% Aromatics in Fuel

E 6 5

0 5

= 4

f 4

;52

>

-

g 3

3

2

2

1

1

0

0

011

0 Vol. %

Water in Fuel

Figure 3. Water tolerance of gasoline blends with ethanol and methanol.

0:2 0.30.40.50.6O h 1:0 Vol. % Water in Fuel

Figure 5. Water tolerance of gasoline blends with methyl alcohol and isobutyl or n-butyl alcohols.

Fuel

in Fuel

0

0.1

Vol. %Water in Fuel

0.2 0.3 0.40.50.60.81.0 Vol. % Water in Fuel

Figure 4. Water tolerance of gasoline blends with isopropyl and methyl alcohols.

direction of the tie lines and show the larger amount of ethyl alcohol cosolvent present in the aqueous phase, as compared to isopropyl and butyl alcohols. The water-tolerance data were plotted on nonlinear graphs and equated to the best fit in the following manner. It was determined that the water tolerance of the fuel, Y4, was a function of the methanol and cosolvent concentrations in the fuel at equilibrium. y4

= f(Y% Y3)

The water-tolerance functions were equated for each cosolvent by graphical techniques: Y4 = exp((bc/S) - a ) - 0.09 (1) where

Figure 6. Water tolerance of gasoline blends with tert-butyl and methyl alcohols.

same equations were used in the computer model to calculate Y4, the water tolerance. Correlations of Data. By necessity, the correlation of experimentaldata and subsequent development of a model are based on the total system composition because the initial tank composition is the only information available in actual operation. Thus, the predicted water bottoms, M , and partition coefficients, Ki, were correlated as functions of tank composition, Ai. Since by material balance Ai = XiM + Yi(100 - M) and

Ki = Yi/X, combining to eliminate Yi gives

a = 1.5 exp(-0.083Y3) + 0.8 b = (0.2(Y3 c = (Yz

+ 1))f

+ 1)8 - 1

and the constants f and g and slope S for each cosolvent are respectively ethyl alcohol 0.78,0.70, 10.7 exp(0.044Y3) - 10; isopropyl alcohol 0.65, 0.65,0.79 + 0.36Y3; isobutyl and n-butyl alcohols 0.57, 0.57, (0.625 exp(-0.1489Y3)+ 0.2)-'; tert-butyl alcohol 0.80,0.61, 10.8 exp(0.027Y3)- 10. The above equations were employed to generate the family of water-tolerance lines for gasoline-methyl alcohol blends that contain the cosolvents ethyl alcohol (Figure 3), isopropyl alcohol (Figure 4), isobutyl and n-butyl alcohols (Figure 5 ) , and tert-butyl alcohol (Figure 6). The

Ai = Xi(M + Ki(100 - M))

(2)

M and Ki are determined by development of suitable regression equations, and Xi is determined by solution of Xi = A i / ( M + Ki(100 - M)) (3) The overall material balance, Bi, is calculated in the usual manner for each component. Bi = (XiM + Yi(100 - M ) ) / A i (4) The calculated values of Xi, Yi, and M are acceptable if Bi = 100 1 vol %. Arbitrary equations were devised and tested in the multicomponent nonlinear regression program for each cosolvent. Due to the synergistic nature of methanol-co-

626

Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985

Table 11. Correlation Coefficients and Equation Numbers for the Regression Equations alcohol ethyl isopropyl tert-butyl isobutyl n-butyl Correlation Coefficient M , high cosolvent 0.992 0.983 0.998 0.981 0.982 M , low cosolvent 0.985 0.975 0.971 0.965 K , gasoline 0.945 0.971 0.954 0.958 0.950 K , cosolvent 0.991 0.960 0.962 0.921 0.944 K , MeOH 0.982 0.973 0.958 0.960 0.972 Average Correlation Coefficient = 0.968

M , high cosolvent M , low cosolvent K, gasoline K , cosolvent K , MeOH

Equation No. 1 2 3 1 I 5 6 6 4 4

2 3 7 6 4

2 3

5 6 4

I

Tank Composition M, v% Water

f

Bottoms

2 3 5 5 4

solvent blends, the equations included cross-product as well as exponential terms. Lengthy equations with seven to eight coefficients proved to be necessary in order to reach a satisfactory fit with experimental data.

Correlations of Data Two types of regression equations were used for M , the water bottoms size, to improve reliability. The first set (M5) was directed toward a high ratio of cosolvent to methanol and the second set (M6) toward a low cosolvent/methanol ratio. The final M was weighted to reflect the average value of the M5 and M6 calculations. This technique was not necessary for tert-butyl alcohol, and only one equation was used. Note that similar results can be obtained with longer, generalized regression equations with 10 coefficients such as 3

f = a.

3

+ ;=1j=1 CCaijAi

(5)

or 3

f =a

3

+ +1j=1 C C a i A i ( l + ajAj)

(6)

Computer Model. Computer logics, Figure 7 , were developed for a cyclic process that simulated gasoline storage conditions at a service station. To be successful, the model was required to calculate sequentially the water bottoms layer, partition coefficients, and gasoline-phase and water bottoms phase compositions that yield good material balances for the components. For this reason the model was fitted with adjustment routines to force compliance with the material balance equation by successive approximations. Water concentrations were calculated from the more reliable graphical water-tolerance equation rather than by using the water K by regression. Another special routine was devised to control the orderly transition of the water bottoms layer from a positive value to zero. Other straightforward tasks of the model included calculating total tank compositions, dispensing gasoline, refilling with portions of fresh gasoline, and recognizing the end point where the water tolerance of the gasoline phase exceeds the total water in the tank as the water bottoms approaches zero. The computer program was written in BASIC language and run on the Wang PCS I1 personal computer and printer. Program Description. The input data is first normaliied to the total tank composition, and the size of water

Final

Yes

KS

New Tank Composition

1

Figure 7. Logics of computer program.

bottoms at equilibrium is calculated from the regression equation. Similarly, partition coefficients Ki are calculated for cosolvent, methanol, and gasoline from regression equations and for water from the water-tolerance equation. At the same time, the compositions of both phases are calculated. A comparison of these values with total tank composition is made to determine the overall material balance of the tank. If the balance is off by more than 1%, small adjustments are made to the above water bottoms and partition coefficients until the balance is 100 f 170. A full printout is made, and a specified portion of the gasoline phase is removed and replaced with fresh fuel. A new total tank composition is calculated, and the above process is repeated. At some point as the tank water content is gradually reduced, the water tolerance of the gasoline phase proves to be greater than the tank water content. The program then exits to an interpolation subroutine for the total number of refills needed to absorb the water bottoms layer. Calculation of the total tank composition involved simple volumetric proportions. Initially and after each refill = ( ~ ~ ( 1 -0 M 0 ) ( N~ yi))/104 + A ~ ’

(8)

As the tank composition approaches complete solubility and the water bottoms approached zero, the precision of the computer model is not great enough to prevent premature extinction of the water bottoms. A logical set of adjustments were made to ensure the smooth transition to zero bottoms for the five types of situations in sequential order. Condition 1. If M > 0.3 vol % , continue main calculations. Condition 2. If E = 1, M = A , - 0.5N4, continue main calculations.

Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985 627

Condition 3. If A4/Y4/ < 3, M = Mi(A4- Yi)/(A; - Yl), continue main calculations. Condition 4. If E = 2 or 3, M = 0.7M1,continue main calculations. Condition 5. For E > 3, M = O.8M?/M2, continue main calculations. A simple relation was found that will predict the number of refills required to dry out the system before the full computer program is started. This calculation of P is used to screen various combinationsof methanol, cosolvent, and water for a target number of refills. p = (100/Gi)((A, - N4)/(y4 - N4)) (9) A rating system for the cosolvent is proposed that is based on the ratio of water tolerance of the fuel to the size of water bottoms. The water tolerance is also related to the alcohol content and water bottoms to the water content of the tank. C = 1000A4(Y4 - O.Ol)/((A,(M + 1))(A2 + AJ) (10) For a given tank composition, the rating C increases with the higher water tolerance and decreases with a higher water bottoms. The main computer program for ethanol-methanol is easily converted for use with the other cosolvents by inserting a subprogram for each of the cosolvents as needed. Operation of the Computer Model. The computer model was programmed to accept as inputs the fresh fuel composition, an initial wate bottoms size, and size of the fresh fuel refill. The model calculates and prints out for each refill the composition of the equilibrium phases, partition coefficients, size of water bottoms, and material balance verification. At the end, when the last of the water layer is soluble in the fuel, a message gives the final tank composition, number of refills, and the cosolvent rating. Various options are available to make changes in the fuel composition and percent refill during a run, to calculate single-stage equilibria, and to shorten the amount of printout information as desired. Full-matrix printouts are given as needed to inspect any of the main parameters. A computer sheet of a typical gasoline storage tank operation for tert-butyl alcohol-methyl alcohol is presented in Table 111. Comparison of Model with Multiple Extractions. Computer predictions of water bottoms growth and extinction were compared with the extraction data from Hutchinson (1982) and from this laboratory. Twelve sets of curves (a-1 in Figure 8) show that the calculated/experimental curves were similar in size and shape and the differences may be attributed to errors in regression or experiments and also to changes in aromatic component distributions. Calculated root mean square differences of predicted vs. experimental were 1.44% for the maximum water bottoms and 0.57 refills needed to dry out the system. Discussion of Results This investigation has developed a computer model that is based on solubility and extraction data for gasolinemethyl alcohol-water mixtures with various cosolvents, including ethyl, isopropyl, n-butyl, isobutyl, and tert-butyl alcohols. For a comparison of the cosolvent efficiency, computer runs were made for each cosolvent with 1.0 vol % water content, 75% tank refill, and 3.5 wt % total oxygen. The results seen in Figure 9 show that ethyl alcohol is a poor cosolvent and that isopropyl alcohol is better but less effective than the three butyl alcohols. Since tert-butyl alcohol produced the lowest water bottoms, it was judged to be the optimum cosolvent. The rating system confirms these visual indications and should

Table 111. Typical Computer Prediction CASOL t-BUTANUL

01/10/85

NO.

110

F h t D l L l t U NO. OF R E F I L L 5 = 3 . 7 9 I

*

*

I

"

*

*

*

.

* *

*

50 RUN NO. 110 I N I T I A L WATER BTnS 0.450% CASULINt t-MUTANUL MEUH FRESH FUEL VOLZ 90.925 4.500 4.500 BUTTOMS CUMP. ' 0.000 0.000 0.000 ThNK CUMP. " 90.333 4.410 4.470 D I S T R I V . COEFF. 16.407 0.154 0.054 2.88Y UPPEK PHASE UDLZ 93.070 3.827 LOWER PHASE " 5.472 24.393 5S.3V4

9 % ALCOHOL K E F I L L NO. 1

*

01/10/85 7 5 . 0 % 1ANK K E F I L L I N L WAICK

0.075 100.000 0.724 0.012 0.213 16.538

9% ALCOHUL K E F l L L NU. 2

50 HUN NO. 1 1 0 01/10/85 XNXTIAL WAltH BlMS O . L h O % 7 5 . 0 % IANK R L b I L L l N t i GASOLINL t-MUlRNUL MEUH WA'TER FRESH FUEL VOLZ 90.926 4.b00 4.500 0.07s TANK C O W . " 88.775 4.95Y 5.640 0.624 D I S T R X I . COEFF. 8.272 0.180 0.070 0.&6 UPPEK P w s t VULZ ~1.566 4.284 3.874 0.270 LUWER PHASE " 11.048 23.701 54.158 10.4/1 EL?UIL. I U I l U M 5 = 3 . 4 4 9 % SUMnAllON = Y/.2---.' YY.Y CYCLtS = 2 MRT'L BALANCES- t i A S . 1 99.V: COSOLV.= 1 0 0 . 0 ; MeOH = 100.0; WATEH = 100.0. COSULVCNI K A I I N L = 1 . 2 0 2 ~~

.

I

*

*

X

.

*

*

*

*

X

*

*

Y% ALLUHUL K E F I L L NU. 3

50 hUN NU. 11U I N X l I A L WAlkh UTM5 0.650% GA5OLINt t-UUIANUL MEOH FRESH FUEL VULZ 90.Y25 4.500 4.500 TANK COW. " 86.311 5.114 4.0YZ D l S T R I U . LUEFI.. 5.009 0.211 0.088 u w t K rwst VOLZ v0.305 4.457 4.73/ 21.914 53.820 LUWEH F H A S t " 18.025 E t u K - w r r u n s = SOX s u n n A i i o N = ~ 7 . 2 - - - , 1 0 0 . 2 MAT'L BALANCLS- G A S . = 100.0; CUSOLU.= YV.9; neUH = COSOLVENl R A l I N G = 1 . 2 3 4

*

*

*

*

*

*

*

*

01/1o/u5 75.0% lANI: K t b I L L I N G WAltH 0.075 0.482 0.051 0.319 4.23Y CYCLES = 2 9V.Y; WAltK = Y Y . Y

* "

WATER TULERANCE EXCEtUS IRNK WAlEH SINGLE GASULINE PHASE I N 'TANK 3 . 7 1 R E F I L L S NEtUEV F O R M l S C l B X L l T Y TANK CUMPUSIIlUN A I 4 I U L L K t F l L L b GbSOLlNt t-MUIANUL VOLY

88./6

b.0140

ntUH

WATtH

b.Yl8b

0.5044

A V C K A C t LUSOLUtNl H A I I N G = 1 . 2 9 6

be useful for further comparisons if other cosolvents are to be evaluated. Keller (1978) reported the effectiveness of cosolvent alcohols to increase with chain length up to at least C6,to decrease with chain branching, and to be greater for primary than for secondary or tertiary alcohols. However, in our analysis of the Cz-C4 alcohols we conclude that effectiveness of cosolvent alcohols increases with chain length but is approximately the same for alcohols of the same carbon number with relatively no dependence on the shape of the molecule or on the location of the alcohol group. Longer chain alcohols may be better cosolvents due to their greater compatibility with gasoline, but eventually these suffer from the gradual reduction in water solubility and functionality that would require higher concentrations in the fuel. A series of model mixtures were run in the computer with tert-butyl alcohol-methyl alcohol blends to test the effect of the important variables on water bottoms growth and the number of required refills. Each of the following variables was evaluated in turn, while keeping the other parameters constant: ratio of tert-butyl alcohol to methyl alcohol, total alcohol concentration, soluble water in the fuel, and the initial water bottoms size. The results in Figure 10 show the large changes created by reversing the tert-butyl alcohol-methyl alcohol ratio from 2/7 to 7/2. Note that the 1:l ratio gives a modest water bottoms growth and small number of required refills. Figure 11 shows clearly the fact that low total alcohol in the fuel extends the number of required refills, while high total alcohol produces the higher water bottoms. A compromise of 9 vol % total alcohol appears to be well justified. As expected, the results in Figure 12 show the appreciable increase in the number of refiis as the soluble water in the fuel is increased but little effect on the water bottoms growth. A symmetrical set of curves seen in Figure 13 was obtained as the size of initial water bottoms was increased. These produced large increases in both the water bottoms

628

Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985

(

D

L

D

b

"

-

O

I

- - . - 0

-m

(3: - N

--

O "

m

O

N

L

r

D

~

L

D

o

u)

L

.*.*. ........

w

...*'>

% S E

...........y................

wj -G 33

(D

m

d

a*..

I m

l "

n

o 7

... r

N

n -

N

o

r

n

o

'*e..

..........

0

?

E

D

r

0

Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 4, 1985 12 11

-

lo

-

Ethanol

1 :1 v CosolventIMethanol O.O5v% Water in Fuel 0.80v% Initial Water Bottoms 75v% Tank Refill Cosolvent - v% Ethanol Isopropanol Iso-n-Butanol 1-Butanol

9-

629

Charge: 1:lvlv Ratio t-Butanol - Methanol O.O75v%Soluble Water in Fuel 0.8v% Initial Water Bottoms 75v% Tank Refill

6 $ 5 vi

3.87 4.28 4.58 4.58

$ 4

2 * 2 3

(0

2“ - 2

ij 7 -

3 1

-E,

R

m

L

0

0

4

0

1

3

2

4 5 6 . 7 Number of Refills

8

9

Figure 11. Effect of total alcohol concentration on size of water bottoms and number of refills.

61 9vi

Charge: 4.5v% t-Butanol 4.5v% Methanol 0.8v% Initial Water Bottoms 75v% Tank Refill

Soluble Water in Fuel, v%

5-

5

t: 4 0

m Number of Refills

Figure 9. Comparison of cosolventa, vol % water bottoms vs. number of tank refills at 3.5 wt % oxygen in fuel.

k

3-

5

2-

b

0

8

iij

10

Charge: 9 . 0 ~ Total 9 ~ Alcohol, O.O75v% Soluble Water in Fuel 0.8v% Initial Water Bottoms, 75v% Tank Refill

1-

0

I

0

8 9[

2

,

I

3 4 5 6 7 8 9 Number of Refills Figure 12. Effect of soluble water in fuel on size of water bottoms and number of refills.

Ratio of t-Butanol To Methanol, vlv

1



Charge: 4Sv% t-Butanol 4.5~96Methanol O.O75v% Soluble Water in Fuel 75v% Tank Refill Size of Initial

I

0

1

2

I

I

3

4 5 6 7 0 9 Number of Refills Figure 10. Effect of tert-butyl alcohol-methyl alcohol ratio on size of water bottoms and number of refills.

growth and number of refills needed. Conclusions Detailed computer studies of model fuels with variations of the important parameters have led to the following conclusions: (1)Neat methanol blends in gasoline require a nearly dry system to prevent the rapid growth of water bottoms. (2) The best cosolvent choice is a compromise between one with high water solubility and one with high gasoline solubility that would increase water solubility in the fuel

Number of Refills Figure 13. Effect of size of initial water bottoms on equilibrium water bottoms and number of refills.

and decrease the alcohol partition into the water bottoms layer. (3) Ethyl alcohol is a poor cosolvent and isopropyl alcohol is a fairly good cosolvent for methyl alcohol/gasoline blends. (4) The butyl alcohols are excellent cosolvents, with tert-butyl alcohol being the optimum.

Ind. Eng. Chem. Prod. Res. Dev. 1905, 2 4 , 630-635

630

(5) The number of required tank refills to dryness can be lowered by reducing the soluble water in the fresh fuel, particularly the water associated with the cosolvent. (6) Problems with water bottoms growth in a tank can be minimized by removing much of the original water bottoms from the tank before the first filling with alcohol/gasoline blend. Acknowledgment I appreciate the assistance provided by the technical staff at the Harvey Technical Center, ARCO Petroleum Products Co. I also thank Karen Meyers for help in the preparation of the manuscript and the management of the ARCO Chemical Co., for the encouragement and permission to publish this work. Nomenclature Ai = tank composition, vol % A[ = tank composition for previous refill, vol % A4 = water in tank, vol % A,,' = water in tank for previous refill, vol % Bi= material balances of components,vol % C = effectiveness rating of cosolvent E = refill number G, = size of fresh-fuel refill, vol % Ki = partition coefficients of components M = water bottoms at equilibrium, vol % M I = water bottoms, first previous refill, vol % M 2 = water bottoms, second previous refill, vol % M5 = water bottoms calculation for high cosolvent in fuel Ms = water bottoms calculation for low cosolvent in fuel Ni = fresh fuel composition, vol % N4 = water in fresh fuel, soluble, vol %

P = predicted number of refills V , = initial water bottoms size, vol % Xi = water bottoms composition at equilibrium, vol % Yi = gasoline-phase composition at equilibrium, vol % Yz= cosolvent in gasoline phase, vol % Y3 = methanol in gasoline phase, vol % Y4 = water tolerance of fuel, vol % i = component (gasoline, cosolvent, methanol, water, respectively) a, b, c, f, g, S = constants and slope for water-tolerance equations Registry No. HzO,7732-18-5;methanol, 67-56-1; ethyl alcohol, 64-17-5;isopropyl alcohol, 67-63-0;n-butyl alcohol, 71-36-3;isobutyl alcohol, 78-83-1; tert-butyl alcohol, 75-65-0.

Literature Cited Cox, F. W. "Component Relationships Within the Two-Phase Gasoline-Methanol-Water", Report 1979, BETC-R1-78/6. Haq, M. A. Hydrocarbon Process. May 1981, p 159. Hutchinson, D. A. ARCO Petroleum Products Co., Harvey, IL, private communication, 1982. Keller, J. L. "Methanol Fuel Modification for Highway Vehicle Use, Final Report", July 1978, NITS HCP/W3683-18. Roehm, P. German Patent 2826883, Dec 29, 1977. Ruiz-Bevia, F.; Prats-Rlco, D. Fluid Phase Equilib. 1983, IO, 77.

Received for review January 22, 1985 Accepted May 20, 1985 Supplementary Material Available: The experimental results of partition coefficients and water bottoms for gasoline, methanol, and water blended with each cosolvent, the regression equations, and the computer model with inserts (17 pages). Ordering information is given on any current masthead page.

Diethanolamine (DEA) Degradationt under Gas-Treating Conditions Chang S.

H s d and C. J. Kim'§

Exxon Research and Engineering Company, Clinton Township, Annandale, New Jersey 0880 1

The degradation of diethanolamine (DEA) under gas-treating conditions was studied by mass spectrometric analysis of the products. I n addition to the formation of 3-(2-hydroxyethyl)oxazolidone-2 (HEO), N,N'-bis(2-hydroxyethyl)piperazine (HEP), and N,N,N'-tfis(2-hydroxyethyi)ethyienediamine (THEED), as reported in the literature, DEA degradation leads to significant amounts of triamine derivatives that were identified as 3-(2-(bis(2-hydroxyethyl)amino)ethyl)-2-oxazolidone (HAO), N-(2-(N,N-bis(2-hydroxyethyl~mino)ethyl~N'~2-hydro~e~yl)piperazine (HAP), and N,N,N",N"-tetrakis(2-hydroxyethyJ~iethylenetriamine (THEDT). A comprehensive reaction mechanism that can account for all the products is proposed and discussed.

Introduction Diethanolamine (DEA), or bis(2-hydroxyethyl)amine, is widely used for removing carbon dioxide, hydrogen sulfide, and other acidic components from gases. During use in plants, however, DEA loses its gas-treating activity as it degrades, causing considerable economic losses and operating problems. DEA degradation yields a complex mixture of polar, high-boiling organic materials that are difficult to isolate for analysis. In 1956, Polderman and Steele found that N,N'-bis(2-hydroxyethyl)piperazine (HEP) is one of the The term "degradation"refers to a decrease in gas-treating activity rather than to a breakdown of the compound. This usage is common in the gas-treating community. Corporate Research-Analytical Sciences Laboratory. f Corporate Research-Science Laboratories. f

0196-4321/85/1224-0630$01.50/0

DEA degradation products and suggested 3-(2-hydroxyethy1)oxazolidone-2 (HEO) as its probable precursor. Hakka et al. (1968) later isolated and identified another degradation product, N,N,N'-tris (2-hydrox yethyl)ethylenediamine (THEED). In addition to these compounds, the formation of high molecular weight compounds becomes important in the later stages of DEA degradation (Kim and Sartori, 1984). These high molecular weight products are very difficult to analyze by conventional GC using a packed metal column because of their high polarity. The study reported here was undertaken to characterize these unknown components with the goal of fully elucidating DEA degradation chemistry. Experimental Section Materials and Sample Preparation. DEA and HEP were purchased from Aldrich Chemical Co. H E 0 and 0 1985 American

Chemical Society