Kato, Y., Koizumi, K., J . Electrochem. Assoc. Japan 2, 309 (1934). Kato, Y., Sugino, K., Koizumi, K., Kitahara, S., Electrotech. J . Japan 5, 45 (1941). Kitahara, S.,Osuga, T., J . Electrochem. Assoc. Japan 10, 409 (1942). Mathers, F., Trans. A m . Electrochem. SOC.17, 261 (1910). Mizuguchi, J., J . Electrochem SOC.Japan 17, 258 (1949a). Mizuguchi, J., J . Electrochem. Soc. Japan 17, 294 (194913). Narasimham, K.C., Sundararajan, S., Udupa, H . V. K., J . Electrochem. Soc. 108, 798 (1961). Osuga, T., Sugino, K., J . Electrochem. SOC.104, 448 (1957). Sampath, S., Thangappan, R., Nachiappan, S. P., Indian Patent 105,731 (Nov. 15, 1967). Schumacher, J. C., Stern, D. R., Graham, P. R., J . Electrochem. Soc. 105, 151 (1958). Sugino, K., Bull. (:hem. SOC.Japan. 23, 115 (1950). Sugino, K., Shibazaki, Y., J . Electrochem. SOC.Japan 16, 9 (1948).
Sugino, K., Yamashita, M., J . Electrochem. SOC.Japan 15, 61 (1947). Thangappan, R., Krishnamoorthy, B., Sampath, S., unpublished data, 1965. Thangappan, R., Krishnamoorthy, S., Sampath, S., unpublished data, 1966. Thangappan, R., Nachiappan, S. P., Sampath, S., Proceedings of International Symposium on Fluidization, Eindhoven, Netherlands, June 1967. Udupa, H. V. K., Narasimham, K. C., Indian Patent 66,195 (Dec. 22, 1958).
RECEIVED for review February 4, 1970 ACCEPTED July 16, 1970 10th Seminar on Electrochemistry, Karaikudi, India, November 1969.
Statistical Concepts in Testing of Dispersants Charles C. Nathan and Clarence 1. Dulaney Betz Laboratories, Inc., Treuose, Pa. 19407 Effectiveness of various surfactants for use as hydrocarbon process antifoulants was investigated by measuring their ability to disperse carbon black particles in kerosine. Stability of the dispersions could be measured reproducibly for materials that were very effective or very ineffective. Systems of intermediate stability gave poorly reproducible data; however, the scatter of the data can be calculated and follows the statistical relations reported previously for adsorption type corrosion inhibitors. The standard deviation, u, of the dispersant efficiency, e, i s expressed by the equation u = 1.33 t(l
-
6).
where
6
This i s the relationship reported for reproducibility of corrosion inhibition data represented the efficiency of inhibition.
C o r r o s i o n and fouling of equipment are problems of increasing technological and economic importance as refinery processes increase in complexity and demands for maximum on-stream time are made on process equipment. The magnitude and interrelation of corrosion and fouling problems are described by Bregman (1963a) in his review of petroleum refining product problems. He points out that a t temperatures above 450"F, corrosion inhibitors lose effectiveness and the use of antifoulants becomes necessary. Bregman (1963b) also discusses the necessity of good dispersant action of an antifoulant additive in order to act on insoluble breakdown products from the processing stream, so that these products are dispersed into the stream rather than adhering to heat transfer surface. A number of papers and patents, such as those by Gonzalez (1969) and Gonzalez et al. (1965), have elaborated on these concepts. As with corrosion inhibitors, numerous compounds and proprietary materials have been developed to remedy refinery fouling problems. Also, as with corrosion inhibitors, it is highly desirable to utilize laboratory screening tests for evaluation of alternative materials before applying them in plant tests. Nathan and Eisner (1958), Nathan (1962), and Carter and Nathan
(1969) have described such tests for screening corrosion inhibitors and pointed out the difficulties of obtaining reproducible results and the necessity of applying statistical concepts in the treatment of experimental data. Nathan (1969) indicates that statistical concepts in the evaluation of organic corrosion inhibitors are also applicable to other systems utilizing surface-active materialse.g., emulsifiers-demulsifiers and dispersants such as those utilized as antifoulants for petroleum processing. This paper gives detailed results and data treatment for some dispersant systems. The papers cited discussed the statistics of adsorptioninhibited metallic corrosion, and showed that corrosion inhibition is reasonably reproducible a t either the 0 or the 100% protection level. In the first case, zero protection, little or no inhibitor is present. In the case of complete protection, the surface is overwhelmed by a large concentration of inhibitor. Between the two extremes in the region of partial protection the data tend t o be very erratic. The adsorption of an inhibitor onto a metal surface from a solution is a dynamic equilibrium. With only a partial surface coverage, there is a high probability a t any time that a surface site may be unprotected. In the media used for Ind. Eng. Chern. Prod. Res. Develop., Vol. 9, No. 4,1970
567
accelerated laboratory tests there is always a strong concentration of corrodent to attack the unprotected sites. Other processes that are controlled by adsorption of surface-active agents should exhibit the same sort of behavior as metallic corrosion inhibition. One such process is illustrated by work done in this laboratory in development of a test for comparing the effectiveness of various process antifoulants as dispersants. Process antifoulants in general are dispersants, antioxidants, chelating agents, or mixtures of these three functions, as shown by Gonzalez et al. (1965). It is desirable to test each function separately, since compounds which are primarily antioxidants or chelants may exhibit some dispersant activity. Experimental
Materials. Atlantic Richfield Co. Ultrasene, a highly purified kerosine, was used without further purification as the dispersing medium. Xylene, ACS reagent grade, was used without further purification. Statex F-12, Columbian Carbon Co., was used without further purification. I t has an average particle size of 29 microns, and a specific gravity of 1.80. The dispersants used, listed in Table I, are identified only as to general type of compound. Most are complex, proprietary mixtures commonly used for dispersing carbon blacks in hydrocarbons and other nonaqueous media. Test Method. A stock solution of carbon black for each day’s runs was made by mixing 4 5 by weight Statex F-12 in xylene, using a high speed blender. The resulting mixture contained on the average 0.023 gram of carbon black per ml of mixture. Sufficient carbon black suspension to make 1000-ppm solids was transferred by pipet to a 250-ml volumetric flask, 1000 ppm of dispersant added, and the flask filled to the mark with Ultrasene. At least six replications were run for each dispersant. The resulting mixtures were shaken 10 minutes a t 120 shakes per minute on a table-type shaker. At the end of the 10-minute period, an 8-ml sample from each flask was pipetted into the spectrometer cuvette. The contents
were allowed to settle without agitation for 20 hours. At the end of the 20-hour period, the amount of light transmitted by the dispersion was measured with a KlettSummerson colorimeter, using a Eo. 62 filter. The colorimeter had previously been calibrated with pure Ultrasene as zero absorption. Blanks were run with carbon black-kerosine dispersion without additive. The average value of the blank was 88, as noted in Table
11. At least two separate sets of determinations in triplicate were run for each dispersant, most cases by different operators. Because of the small number of replicates, it is impossible to determine the operator bias, if any. The scale on the Klett-Summerson colorimeter is logarithmic. Thus, the colorimeter reading is a direct function of the number of particles suspended, provided Beer’s law is followed. The average blank, 88, is taken as zero suspension and minimum absorption, and the maximum reading, 950, taken as 1007 suspension and 10Orc absorption of light. The resulting “percentage suspensions” are given in Table 111. Discussion
Ideal Case. Assume that the mechanism of carbon black dispersion is similar to that for inhibition of corrosion by adsorption-type inhibitors-when sufficient additive is present, the carbon black particles are all dispersed, and when no additive is present, none of the particles is dispersed. When an intermediate concentration of dispersant is present, the data should become very erratic. This implies that the action of the dispersant (and the dispersion of the carbon black) is controlled by an adsorption equilibrium, similar to that reported for adsorption-type corrosion inhibitors. Second, assume that the absorption of light by the carbon black dispersion follows Beer’s law.
Table II. Colorimeter Readings for Carbon Black Dispersions Dispersant
A
Table I. Approximate Chemical Composition of Dispersants Used Dispersant
A B
C D
E F
G H J K I, M N P Q
B
C
Composition
D E F G
Substituted succinimideO (Lubrizol 894) Ethoxylated aliphatic amine Aromatic sulfonamide Ethoxylated aliphatic amine Calcium sulfonate Substituted succinimide, Betz Petromeen AF-12 Substituted succinimide, Betz Petromeen AF-104 Substituted succinimide, amine, DETAO Substituted succinimide, amine, Primene JMT‘ Substituted succinimide, amine, Primene J M T Substituted succinimide, amine, Primene J M T Substituted succinimide, amine, Primene 8 1 R Substituted succinimide. amine, DETA‘ Substituted succinimide. amine, TETAd Substituted succinimide, amine, TETAO
’ All succinimides believed to he made by reaction of poly(tertbutyl succinic anhydride) with an amine. Dispersants H, 3, K , L, N, P, and Q have increasing ratio of amine to anhydride. * DETA. diethylenetriamine. Trade marks. Rohm and Haas Co. TETA, triethylenetetramine.
568 Ind. Eng. Chem. Prod. Res.
Develop., Vol. 9, No. 4, 1970
Readings
770, 690, 920, 950, 950, 950, 810, 810, 810 700, 660, 740, 950, 950, 950 620, 630, 690, 324, 450, 385, 680, 610, 610, 950, 950, 950 950, 950, 950, 835, 835, 810. 950, 950 950, 950. 950, 950, 950, 950 950, 950, 950, 478, 600, 590 950. 950. 950. 740. 740, 740
H J
300, 435, 375, 455, 660, 660 315. 317, 320, 680, 521. 521
K
326. 375. 325. 190. 170, 199
L
229, 199, 182. 100, 103, 118 107, 84. 101. 183. 175, 172
M N P
Q Controls
360, 276, 700. 224, 248, 315 284, 284, 284. 950. 950, 950, 475. 470. 410 240, 410, 315, 140, 104, 133, 620, 397 83, 92, 92, 95, 77. 8T
Underlined tests run by operator A. all others by operator
B.
Table Ill. Percentage Dispersion of Carbon Black Dispersont
A B C D E F G
H ?J K L M
N P
4
Oh
Dispersion
79, 70, 96, 100, 100, 100, 84, 71, 66, 75, 100, 100, 100 61, 63, 70, 27, 42, 34, 69, 100, 100, 100 100. 100, 100, 93, 86, 86, 83, 100, 100, 100,100,100, 100 100, 100, 100, 45, 59, 58 100, 100, 100, 75, 7 5 , 75 24, 40, 33, 42, 65, 65 33. 26, 28. 69, 50, 50 27, 33, 12, 9, 13, 27 16, 11. 10, 1, 1, 4 2, 0, 1, 10, 9. 8 31, 22, 71, 16, 18, 26 22. 22, 22, 100, 100, 100, 44, 27, 37, 26, 6, 2, 5, 60,
84, 84
60, 60 100, 100
45, 37 36
where I and I,, are the intensity of light transmitted by the dispersion and the carbon-free solvent, respectively, ii is an effectiveness factor, and c is concentration of dispersed particles in appropriate units. If the above conditions are fulfilled, the relation between the average fraction dispersion, C , and the standard deviation, U, should be given by Expression 2, which has been reported for corrosion data. u
= 1.33 ( t ) ( l - E)
u =
[(x - x - ) L / n- 13‘
(3)
Actual us. Expected Standard Deviation. Table IV gives the average percentage dispersion of the carbon black, the calculated standard deviation, and the “expected” standard deviation calculated by Equation 2. The agreement between the actual and expected standard deviations is good. This is particularly significant, in that so few runs were obtained for each dispersant, and all runs were made in one laboratory. Conclusion
The mechanism presented above is probably correct. This indicates that other processes controlled by adsorption of surfactants from solution may show the same type of behavior as the corrosion and dispersion cases Table IV. Calculated and Expected Value of Standard Deviation of Carbon Black Dispersion Data Av
Yo
Dispersant
Dispersion
ucc
ue‘
A B C D E F G H J K L kl
88 85 65 94 100 77 87 45 39 20
11 16 25 7
12 16 30
0
0 23 14 32 32 21 8 6 28 33 25
S
P
4 ‘ r ~ c=
-
5 31 54 25
25 13 17 21 10 6 4 20 35 20
n
calculated standard devlatlon oe = expected standard d e w t i o n
so far cited. The emulsification of oils by surfactants will be examined in a future paper. If dispersants are to be compared in effectiveness by a dispersion test, they should be compared a t the minimum concentration necessary to give 95% dispersion. If they are compared a t any lower percentage dispersion, the uncertainty of the data will be so great that the results will have little meaning. NOTE ADDEDIN PROOF.The poor reproducibility of the data a t intermediate values of dispersancy has been questioned by a reviewer as being indicative of poor techniques; irrespective of the test method employed, careful laboratory work should give reproducible data. He points out that triplicate determinations by the same operator appear t o be fairly consistent, as compared to differing values obtained with different operators. The authors realize the seeming inconsistency of such large variations in replicate determinations and point out that these variations have been reported previously in other systems where adsorption of surfactants is a controlling mechanism in the behavior of the system. This sort of behavior, with resulting poor reproducibility, has been described by Nathan (1958, 1969), Carter and Nathan (1969), and Nathan and Eisner (1958). Efforts to reduce data spread by refining the systems a n d / o r techniques were unsuccessful, as pointed out by Nathan and Eisner and Nathan. Furthermore, the variations within a given laboratory were as great as those between laboratories for corrosion-inhibition measurements. Similarly, for the dispersancy data reported, a close evaluation of the results shows that the variations between operators are about the same as those in replication by an individual operator. (The data tabulated in Table I1 are given in groups of three replications by the same operator. Those underlined are by Operator I. Others are by Operator 11.) The authors emphasize that systems employing surfactants-e.g., corrosion inhibition, dispersancy, and emulsification-may be expected to give data of poor reproducibility a t intermediate efficiency values, but that the reproducibility is predictable from empirical relations presented and must be considered if valid comparisons are t o be made between alternative materials. Literature Cited
Bregman, J. I., “Corrosion Inhibitors,” p. 257, Macmillan, New York, 1963a. Bregman, J. I., “Corrosion Inhibitors,” p. 279, Macmillan, New York, 196313. Carter, D. A . , Nathan, C. C., M a t e r Prot 8, 61 (1969). Gonzalez, G. A., L.S. Patent 3,442,791 (May 6, 1969). Gonzalez, G. A., Hagney, D. J., Sumner, N. E., “Evaluation and Application of Metal Coordinating Antifoulants,” NACE Southern Regional Conference, New Orleans, La., Oct. 18-21, 1965. Nathan, C. C., Corrosion 18, 285T (1962). Nathan, C. C., “Statistical Concepts in Evaluation of Corrosion Inhibitors and Other Surfactants,” NACE South Central Regional Meeting, Houston, Tex., Oct. 14-16, 1969. Nathan, C. C., Eisner, E., Corrosion 14, 193T (1958).
RECEIVED for review February 17, 1970 ACCEPTED August 12, 1970 Ind. Eng. Chem. Prod. Res. Develop., Vol. 9, No. 4, 1970 569