Estimation of Chemical Incompatibility (Other-Chemical Reactivity) By

ESTIMATION OF CHEMICAL INCOMPATIBILITY {OTHER-CHEMICAL REACTIVITY) BY COMPUTER1

DALE N. TREWEEK, 2626 Chartwell Road, Columbus, OH 43220 CARL A. ALEXANDER, JAMES R. HOYLAND, AL F. FENTIMAN, WILLIAM M. PARDUE, Battelle, Columbus Laboratories, 505 King Avenue, Columbus, OH 43201.

Abstract. Two simple parameters—the National Academy of Science-National Re- search Council "Other Chemical" rating and a Lewis acid-base rating were combined in a linear discriminate model to predict the degree of hazard from binary mixtures as measured by Dow Chemical experimental tests. The observed error of 10% com- pares favorably with 40% error obtained using the NAS-NRC rating system. These results will be incorporated in the American Society for Testing and Materials computer program CHETAH (Chemical Thermodynamics and Energy Hazard Potential).

OHIO J. SCI. 80(4): 160, 1980

The accidental mixing of industrial mishap make some type of model or chemicals during transportation, storage simplification highly desirable. Such a or processing poses a potential energy model might serve to reduce the potential hazard, denned in terms of both the hazard by identifying hazardous combi- relatively immediate release of heat nations, which can then be precluded by and/or gas as a result of chemical in- physical separation with relatively inert compatibility or "other chemical" re- chemicals. activity and the potential formation of an explosive mixture that can subse- Background quently be initiated by an appropriate A review of the more common hazard stimulus (mechanical impact, thermal or rating systems (over 30 have been physical shock). In some cases, the counted) resulted in the summary of circumstances of an accident (collision, reactivity definitions shown below. fire, etc.) causing the inadvertent mixing REACTIVITY DEFINITIONS of some unusual combination of chemicals USCG—(NAS-NRC)—Other Chemical- can also serve as the initiation stimulus Reactivity for a violent deflagration or detonation Based on possible reactivity with of such a mixture. members of other grades Water Presently, there are about 1000 chemi- Based on mixing with equal weight cals transported in bulk in the U.S. of water at ambient temperature Many thousands more are transported Self in smaller quantities. Statistically, the Usually based on polymerization number of combinations by which just 2 Fire Hazard of the bulk-shipped chemicals might be Based mostly on flash points but accidentally mixed is beyond experi- noxious gas production or air/ water spontaneous ignition con- mental evaluation. Chemical tankers, sidered for instance, may contain thousands of DOW—Binary Mixture kilograms of a half dozen or more dif- Based on temperature rise and gas ferent chemicals in relatively close prox- evolution imity. Therefore, the astronomical num- NFPA—Reactivity ber of unusual combinations that might Based on shock sensitivity, re- accidently arise during a transportation activity with water Flammability Based on ignition temperature and iManuscript received 28 March 1979 and in tendency to form explosive dusts revised form 21 January 1980 (#79-20). or mists 160 Ohio J. Sci. CHEMICAL COMPATIBILITY ESTIMATION 161

ASTM—Hazard Potential TABLE 3 Susceptibility to ignition or re- NFPA Reactivity {shock sensitivity and lease of energy under varying en- water reactivity). vironmental conditions Detailed study of these systems reveals a GRADE marked difference in the definition of 0 Stable and unreactive with water. 1 Unstable at high T and P or react terms like "reactivity," a vast difference slightly with water. in the degree of subjectivity required to Unstable, capable of violent chemi- assign a given chemical to a particular cal change but not detonation or re- hazard class, and in some cases, a lump- act violently with water. Shock sensitive but require strong ing of more than one type of hazard into initiating source or high T and P, or the same rating system to facilitate use react explosively with water. by a particular group. Shock sensitive at normal T and P. The diversity of these hazard rating systems relating to chemical compatibility is exemplified by the U. S. Coast Addressing the chemical incompatibil- Guard (National Academy of Science ity or "other chemical" hazard exclu- 1973, Department of Transportation sively, the Dow Chemical binary mixture

TABLE 1 Other Chemical Reactivity Hazard Rating System* {mixing results in overflow or rupture of tanks, ignition or evolution of noxious gases). GRADE 0 UNREACTIVE EXCEPT WITH GRADE 4 Saturated hydrocarbons 1 MILDLY REACTIVE WITH GRADE 4 Aromatic and olefinic hydrocarbons 2 MAY REACT WITH GRADES 3 OR 4 ONLY Alcohols, aldehydes, etc. 3 MAY REACT WITH GRADES 2, 3, OR 4 Organic acids, amines, etc. 4 MAY REACT WITH ALL OTHER GRADES Cone, mineral acids and bases

"National Academy of Science-National Research Council (NAS 1973).

1975), which considers reactivity in 3 rating system (Flynn 1970) developed distinct rating systems: under contract to the National Academy Other chemical (see table 1) of Sciences Advisory Committee for the Coast Guard is very explicit about the Water and Compatibility criteria used to assign a chemical to a of Chemicals (see table 2) particular class (see table 4), whereas the whereas NFPA (National Fire Protec- TABLE 4 tion Assoc. 1973) lumps many of the Dow binary mixture {temperature rise and same considerations into its reactivity gas evolution). system: Reactivity shock sensitivity, de- HAZARD DEGREE composition or poly- 1 NONE AT max. <25 °C, no gas merization, reaction 2 LOW AT max. 25-50 °C, no gas with water (see table 3) 3 MEDIUM AT max 50-75 °C, no gas 4 HIGH AT max >75 °C or gas evolution TABLE 2 Water {mixing with equal weight of water at ambient temperature). NFPA and NAS-NRC systems use un- quantified descriptors like very, normal, GRADE moderate, mild, and vigorous. To quan- 0 No reaction. tify the NAS-NRC "Other Chemical" 1 Mild reaction. 2 Moderate reaction. system, a flow chart to assist in classify- 3 May be hazardous. ing an unknown was developed (figure 1). 4 Vigorous—hazardous. A computerized technique based on this flow chart would require a test for 162 D. N. TREWEEK, ET AL Vol. 80 was tested with a representative Grade 3 chemical (i.e., ammonia), no reaction START potential would be observed (Flynn 1970) and the unknown would pass the Grade 3 test. This same unknown, however, has been observed experimentally by Dow to be highly reactive with another Grade 3 chemical, ethylene diamine (A T-194C, A P-13.5 psi, High Hazard) (Flynn 1970). Herein lies the problem in simplifying binary or other chemical reactivity using the NAS-NRC rating system. The CG Guide to Compatibility of Chemicals (Circular No. NVIC 4-75) provides an ideal working document for field use, but the "GO/NO-GO" simplification that makes this guide so attrac- tive for field use also makes it less suited for quantitative evaluation of a new model. Thermodynamic parameters were found to have little or no power in predicting binary reaction tendency (Alex- ander el al 1975). Considering the low temperatures generally encountered, this finding is consistent with historical ob- servations on the power of such calcula- tions. Although the Dow rating system ap- pears to offer many advantages in terms FIGURE 1. Flow chart to classify an un- of reference utility to a research study known into the NAS-NRC Other Chemi- cal System. (0), (1), means test for re- (Flynn 1970), several irreconcilable dif- action with most reactive member of ferences have been encountered as a re- group 0, 1, etc. R—NAS other chemical sult of Dow's noble efforts to reduce the hazard rating (NAS 1973). amount of experimental data required. Specifically, Dow assigned 209 bulk- reactivity with the most reactive member shipped chemicals to 59 groups, each with of each grade (0 through 4). Some of a designated representative working the grades contain a variety of types of chemical. This reduced the amount of chemicals; therefore, to account for the binary experimental data required from possible wide spectrum of reactivity the 21,736 to 1711. unknown might exhibit because of its Ethylene glycol was selected by Dow own characteristics and because of those as the representative working chemical of the other reactant, the test with the for the alcohols as shown in table 5. In most reactive member of each grade Dow's experimental evaluation of this might conceivably also require testing chemical with 73% nitric acid (another with the most reactive member of each working chemical), no reactivity was ob- type of compound within each grade. served, and thus no reactivity is pre- Otherwise, the unknown might "fall dicted for any of the other alcohols of through" the flow chart with no estimated this group with nitric acid. Unpub- reaction potential (i.e., 0 grade) when in lished experimental data on similar 10 fact, one type of compound not tested ml quantities of several other alcohols might react quite vigorously with the un- of the same Dow group as ethylene glycol, known. For example, if ethylene chloro- however, resulted in no reactivity in 3 hydrin was the unknown (actual NAS- cases (glycerin, tertiary butyl alcohol NRC "other chemical" grade of 2) and and iso-butyl alcohol) and very high, al- Ohio J. Sci. CHEMICAL COMPATIBILITY ESTIMATION 163

TABLE 5 Comparison of binary mixture data with current ratings.

Battelle DOW** Binary Mixture Experimental NAS-NRC* Rating USCG*** Data Guide System NVIC 4-75 Nitric acid (70%)/ethylene glycol Nt H N X Nitric acid (70%)/glycerin N H N X Nitric acid (70%)/butyl alcohol, tert- N H N X Nitric acid (70%)/butyl alcohol, iso- N H N X Nitric acid (70%)/butyl alcohol, sec- H H N X Nitric acid (70%)/isopropyl alcohol H H N X Nitric acid (70%)/cyclohexanol H H N X

*NAS 1973 **Flynn 1970. ***DOT 1975. fN = no reaction, H = hazardous, and X = unsafe combination. most hypergolic, reactivity for 3 others might lead to greater accuracy. Spe- (secondary butyl alcohol, isopropyl al- cifically, the relative strength of the acidic cohol and cyclohexanol). On this basis, (proton donating or electron pair accept- transportation cost penalties may be im- ing) plus basic (proton accepting or elec- posed on safe combinations and, at the tron pair donating)—or Lewis char- same time, potentially catastrophic com- acteristic of the reacting pair was estab- binations may be permitted. In fact, lished using a simple ranking similar to the current Coast Guard chemical com- the National Academy of Science other patibility guide (DOT 1975) has been chemical system. This ranking was observed to permit mixtures with deto- based on known acid-base strengths, nation potential such as nitrobenzene/ bond dissociation energies required to nitric acid (during WW I, nitrobenzene/ nitrogen oxide mixtures were used in air- TABLE 7 craft bombs) (Federoff 1962, Van Dolah Lewis Acid-Base Ratings for chemicals. 1969). REACTIVITY Procedure MINERAL STRONG ACIDS Review of the National Academy of Acetic Acid STRONG Science other chemical rating system Acrolein *\ (NAS 1973) and its limitations resulted Methyl Ethyl Ketone ' MODERATE Butyraldehyde J in the generalization that additional in- LEWIS formation about the reaction tendency ACIDS Acrylonitrile ^ between the two particular chemicals Vinyl Acetate Ethylene Cyanohydrin > WEAK TABLE 6 Styrene ^ Ethylene Chlorohydrin J Relative strength of Acid-Base Reaction Tendency. n-Hexane NEUTRAL

GRADE Isobutyl Alcohol "> 0 Weak Lewis acid/weak Lewis base Ethylene Glycol 1 Moderate Lewis acid/weak Lewis Isopropyl Alcohol, Cyclohexanol I > WEAK base Glycerin, Sec-Butyl Alcohol i Weak Lewis acid/moderate Lewis Tert-Butyl Alcohol base LEWIS Aniline J BASES Moderate Lewis acid/moderate Propylene Oxide "*) Lewis base Ethanolamine • MODERATE Strong Lewis acid/strong Lewis Ethylenediamine J base Strong Lewis acid/mineral base ^ Ethylenimine STRONG Mineral acid/strong Lewis base MINERAL Mineral acid/mineral base VERY BASES STRONG 164 D. N. TREWEEK, ET AL Vol. 80 produce a proton or proton accepting known factors such as free radical param- species, etc. eters, non-polar solvent effects and ionic To develop the acid-base reaction reaction parameters such as heats of solu- ranking, mineral acids reacting with tion, dynamic and static induction, reso- mineral bases were placed at one end of nance stabilization, hyperconjugation, the spectrum and weak Lewis acids or double bond polarizability and ease of bases in the middle, as shown in table 6. formation of carbonium ions should be Specific Lewis acid-base ratings for some considered in future research. Refine- of the chemicals considered are shown in ment of the Lewis acid-base ranking table 7. It was noted that some of the should also be considered (Drago 1973). chemicals in the center (with weak Lewis We combined the Lewis reaction ranking characteristics) were amphoteric; other mathematically with other published

TABLE 8 Vector Classification of Chemical Incompatibility (Flynn 1970) using National Academy of Science "Other Chemical" Hazard Ratings (NAS 1973) and Lewis Acid-Base Reaction Tendency (table 6).

Pattern Other Experimental Recognition Chemical (estimated) Hazard! Degree Class Vector Error* Estimated Error* C1ass++ Class** Ethylene Chlorohydrin/Ethylene Diamine H 4 3 U H — Vinyl Acetate/Ammonia, 28% L 2 1 U H 0 Aniline/Acrolein M o 2 U H 0 Acetic Acid/Caustic Soda, 50% H 4 4 H — Aery loni trile/Ethanolamine H 4 4 H — Styrene/Sulfuric Acid, 96% H 4 4 — H — Propylene Oxide/HCL, 35% H 4 4 — H — Ethanolamine/Acetic Acid H 4 4 H — Aery loni trile ,/E thy lenimine L 2 2 — H 0 Ethylene Chlorohydrin/Caustic Soda, 50% H 4 4 — H 0 Ethylene Diamine/Acrolein H 4 4 H — Butyraldehyde/Ethylenediamine M 3 3 H 0 Butyraldehyde/Ethanolamine M 3 3 H 0 Aniline/Butyraldehyde L 2 2 H 0 Ethylene Diamine/Methyl Ethyl Ketone L 2 2 — H 0 Ethylene Diamine/Ethylene Cyanohydrin L 2 2 H 0 Acetic Acid/Aniline N - 1 1 H 0 E thylene Chlorohydrin/E thanolamine N 1 1 — H 0 Aniline/Ethylene Diamine N 1 1 — H 0 Ethylene Diamine/Ethanolamine N 1 i H 0

N-Hexane/Ethylene N 1i— i 1i— i — N — N-Hexane/Acry loni trile N N — N-Hexane/Styrene N 1 i N — N-Hexane/Acetic Acid N 1 1 — N — N-Hexane/Aniline N 1 1 — N — N-Hexane/Butryaldehyde N 1 1 — N — N-Hexane/Ethylenediamine N 1 i N — N-Hexane/Ethanolamine N 1 1 — N — N-Hexane/Propylene Oxide N 1 1 — N — N-Hexane/Ethylene Chlorohydrin N 1 i N — Totals 3(10%) 12(40%)

*U = underestimated, 0 = overestimated. **H = hazardous, incompatible; N = non-hazardous, compatible. +See table 4, Flynn (1970). ++See table 9. [National Acad. Sci. (NAS, 1973). Ohio J. Sci. CHEMICAL COMPATIBILITY ESTIMATION 165 chemical hazard ratings of the chemicals the higher side of the vector consistent (NAS 1973) using a linear discriminant with DOT product conventions employed. analysis approach of pattern recognition The substantial reduction in error com- (Kowalski 1972). pared with the best rating system avail- able prompts inclusion of this new FINDINGS method into the ASTM computer pro- Chemical incompatibility was esti- gram CHETAH (Seaton et al 1974). mated using only two parameters—the Further improvement of the model will NAS other chemical rating for each com- be required to be consistent with this ponent of the mixture plus the Lewis program's self-reactivity hazard predic- acid-base rating—with about 10% error tions; however the many relevant reac- compared to 40% using the NAS hazard tion parameters not considered should rating system alone (table 8). Since all further improve the demonstrated estima- binary mixtures with a Dow hazard class tion capability for binary mixtures. of 2 or more are considered unacceptable Acknowledgment. We would like to thank (DOT 1975), the performance of this Ralph Wilson for his assistance with our model is better than 10% in terms of computer work and Ann Bertagnolli for rear- field use. ranging the tables.

TABLE 9 Linear discriminant vector* coefficients to classify chemicals in table 8.

a0 a3 Vector for: Classes 1 and 2 2.6994E -3.7935E 4.0874E -5.3298E Classes 1 and 2 1.0666E -2.7525E 1.4428E -7.1032E Classes 2 and 3 6.0000E -3.3333E -1.3333E -1.0000E Classes 1 and 4 1.8688E -1.8702E -1.7661E -6.7407E Classes 2 and 4 3.5823E -1.5190E -5.8228E -8.9114E Classes 3 and 4 6.1048E -8.0645E -8.1452E -9.1129E

NVR *Vector = an+• S a;\ i l Vari or 2 = National Academy of Science "Other Chemical" Reactivity Hazard Rating (NAS 1973) for each chemical and Var3 = Relative strength of Acid-Base Reaction Tendency (see table 6).

The linear discrimanant vector co- LITERATURE CITED efficients used to assign these binary mix- Alexander, C. A., J. R. Hoy land, and D. N. tures with known chemical incompati- Treweek 1975 Evaluation of computerized techniques for predicting chemical reactivity bility hazards are given in table 9. and stability. NTIS A009-561, U. S. Coast These vectors take the form: Guard Rept. 1/13/75. Department of Transportation (DOT), U. S. NVR Coast Guard 1975 Navigation and Inspec- VECTOR = ao+ S aiVARi tion Circular No. 4-75, Washington, D.C. 20590. Drago, R. S. 1973 Structure and bonding— quantitative evaluation and prediction of where: NVR = number of variables donor-acceptor interactions. Springer-Ver- a = coefficients published lag, Vol. 15. VAR = variables in order listed. Federoff, B. T. 1962 Encyclopedia of Explo- sives and Related Items. Vol. 2, Picatinny In order to classify additional unknowns, Arsenal, Dover, NY. Flynn, J. P. 1970 Classification of chemical one need only compute the sign of reactivity hazards. CCT-40-69-15, National VECTOR; positive values lie on the lower Academy of Sciences/U. S. Coast Guard class side of the vector and negative on Rept., 12/70. 166 D. N. TREWEEK, ET AL Vol.

Kowalski, B. R. and C. F. Bender 1972 Pat- Amer. Soc. Testing and Materials, Phila- tern recognition. A powerful method for delphia, PA 19103. analyzing chemical data. J. Amer. Chem. Treweek, D. N., J. R. Hoy land and C. A. Soc. 94:5632. Alexander 1975/1976 Use of Simple Thermo- National Academy of Science 1973 Evaluation dynamic and Structural Parameters to Pre- of the Hazard of Bulk Water Transportation dict Self-Reactivity Hazard Ratings of of Industrial Chemicals. Washington, D.C. Chemicals. J. Hazardous Mater. 1: 173-189. , W. M. Pardue, J. R. Hoyland, C. A. National Fire Protection Association (NFPA) Alexander, W. H. Seaton and E. Freedman 1973 Fire Protection Guide on Hazardous 1978 Estimation of Explosive Hazard by Materials. NFPA, 470 Atlantic Avenue, Computer. Ohio J. Sci. 78: 245-254. Boston, MA 02210. Van Dolan, R. W. 1969 Detonation potential Seaton, W. H., E. Freedman, and D. N. Tre- of nitric acid systems. AICHE Symposium week 1974 CHETAH—The ASTM Chemi- on Loss Prevention in Process Industries. cal Thermodynamic and Energy Release 64th Natl Meeting, New Orleans, March 16- Evaluation Program, DS-51, 05-051000-31, 20. Preprint 26B, Pt. 2.