crw 2.0: a representative-compound approach to functionality-based prediction of reactive chemical...

CRW 2.0: A Representative-Compound Approach toFunctionality-Based Predictionof Reactive Chemical HazardsLewis E. Johnsona and James K. Farrba Department of Chemistry, University of Washington, Seattle, WA 98195b Hazardous Materials Response Division, National Oceanic and Atmospheric Administration, NOAA Western Regional Center,Seattle, WA 98115; [email protected] (for correspondence)

Published online 19 February 2008 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/prs.10248

The NOAA Chemical Reactivity Worksheet (CRW)has provided a user-friendly, yet powerful method ofqualitatively predicting reactive chemical hazardssince its introduction in 1998, and has seen its usein spill response, storage management, and processsafety for intentional chemistry. The CRW predictsreaction hazards based on a database of more than6,000 common industrial chemicals, classified bymajor reactive groups, using 43 3 43 matrix of or-ganic and inorganic reactive groups, modified fromthe earlier ‘‘EPA method for determining the compati-bility of hazardous mixtures.’’ CRW output describesthe hazards qualitatively; e.g., ‘‘spontaneous ignitionof reactants or products due to reaction heat,’’ or‘‘combination liberates nonflammable, nontoxic gasand may cause pressurization.’’ The CRW also pro-vides summaries of properties and special hazards formany of the compounds in the database.

CRW 2.0, a new, standalone version of the CRW, isdue to be released later this year. Earlier versions ofthe CRW had a limited set of references and limitedinformation about specific reactions, and this infor-mation could only be found in each compound’schemical profile, increasing search time for assessing

reaction specifics, and providing little informationthat could be used to mitigate the chance of false-pos-itives—the CRW is designed to be conservative in itspredictions, and will indicate an incompatibility evenif only a few reactions are found between two func-tional groups. Version 2.0 improves the databasebased on an extensive literature survey of potentiallyhazardous reactions for representative members of all43 listed functional groups, with primary literaturereferences available for many of the hazards pre-dicted, and many errors and omissions corrected.This article describes the methodology for the litera-ture search, rationale for changes to the operation ofthe database, extent of the new documentation,and the remaining design limitations of the product.� 2008 American Institute of Chemical EngineersProcess Saf Prog 27: 212–218, 2008

Keywords: reactivity, reactive, hazards, compatibility

INTRODUCTIONSince 1998, the NOAA Chemical Reactivity Work-

sheet (CRW) [1] has provided an easy-to-use methodof predicting incompatibilities because of inadvertentmixing of common chemicals, and hazardous reac-tions of industrial chemicals, providing informationon the interactions of over 6,000 common industrialcompounds and mixtures. It has found applicationsin prediction of storage compatibility, intentionalchemistry/process safety, and response to hazardouschemical release incidents. It has been implementedboth as a stand-alone product, and as a part of the

James K. Farr is the principal investigator. (e-mail: [email protected])This work was supported by Center for Chemical Process Safety (CCPS);

Technology Alliance of the American Institute of Chemical Engineers(AIChE); The National Oceanic and Atmospheric Administration (NOAA).

� 2008 American Institute of Chemical Engineers *This is a U.S.Government work and, as such, is in the public domain in theUnited States of America.

212 September 2008 Process Safety Progress (Vol.27, No.3)

CAMEO suite of emergency response and planningsoftware.

The CRW bases its reactivity predictions on a two-dimensional matrix containing 43 different organicand inorganic reactive groups, a classification forinorganic compounds that is not described by any ofthe other groups, nine special hazards (e.g. polymeri-zation, explosive, water-reactive, etc.), a nonreactiveclassification, and an unknown classification. Thespecial hazards are based on a set of special hazardsdeveloped for the EPA method for determining thecompatibility of hazardous mixtures [2] (EPA Table),although additional hazards such as air-reactive andperoxidization have been added; these groups areintended to inform the user when a compound hasparticularly high intrinsic reactivity. The ‘‘unknown’’group is intended to list when a compound cannotbe classified using the stated reactive groups, sinceinsufficient classification information is provided in aparticular mixture’s department of transportationrecords, or where composition of a mixture mayvary. Compounds are assigned to one or more reac-tive groups based on chemical structure and function-ality, and binary pairs of individual functional groupsare used to determine reaction hazards [1]. Such anapproach provides first-order predictions withouthaving to determine hazards for every pair of materi-als in the database. Determination of hazards for ev-ery pair would require up to 18 million individualentries, many of which have little or no availablereactivity information. Additionally, extensive rework-ing of the product would be needed each time a newcompound was added.

While the reactive-group based approach can pro-vide useful predictions with negligible computationalcost, earlier versions of the CRW provided insufficientdocumentation about how a pair of functional groupswas determined to be incompatible. Some literaturecitations could be found under each compound’s‘‘chemical profile,’’ a section of a compound’s entryabout properties and special hazards [1], but therewas no direct access to citations from the reactivehazard report generated by the CRW and no cross-referencing between compounds, increasing the diffi-culty of accessing the documentation.

The latest version of the CRW, version 2.0, whichis pending release as a stand-alone FileMaker data-base, addresses the documentation issue by providingliterature citations for hazardous reactions of re-presentative compounds for each reactive group,arranged as hazard statements for each binary pair ofgroups. These extended hazard statements are nowavailable for the majority of binary pairs and provideguidance to users in interpreting the CRW’s predic-tions. The stand-alone product will also incorporateseveral other new features including the ability forusers to create custom databases of compounds thatthey are using that are not included in the NOAAdatabase. The compatibility table used in CRW 2.0will also be accessible through NOAA’s new CAMEOchemicals web portal [3], but currently lacks theimproved documentation; most CRW 2.0 featureshave not yet been implemented online.

REPRESENTATIVE COMPOUNDS APPROACHThe 43 functional groups selected for the CRW are

intended to cover a wide range of both organic andinorganic reactions to maximize the scope and accu-racy of the CRW’s predictions. Many representativeindustrial compounds can be found for each group.However, those compounds can produce differentdegrees of hazards because of factors beyond the tar-get functional group. The process given laterdescribes how representative compounds were cho-sen for the literature search used to overhaul theCRW’s predictive matrix.

Compounds within each group can have varyinglengths of hydrocarbon chains, amounts of branchingor cyclization, degrees of substitution, levels of conju-gation for unsaturated compounds, or counter-ionsfor salts. These structural differences can affect kineticbehavior at reaction centers, making generalization ofhazards or lack of hazards to an entire reactive groupnontrivial. Some of these structural differences, suchas those between primary, secondary, and tertiaryalcohols and amines, may warrant further study of ki-netic effects, and possible revision of the reactivegroups.

Classification is further complicated by many com-mon industrial compounds possessing multiple func-tionalities. For example, the CRW classifies acroleinas both an aldehyde and an unsaturated hydrocar-bon, and hydroxylamine as both an amine and astrong base [4]. If a compound is multiply classified,the group suspected to contribute most to the com-pound’s reactivity is listed first in that compound’sCRW entry. However, both reactive groups are addedto the binary mixture, potentially introducing errorfrom either overcounting or neglect of activation orinhibition by other groups.

In order to alleviate the multiple-classification di-lemma, monofunctional compounds were chosen asrepresentative for a functional group, since most ofthe multifunctional compounds in the database areassigned to multiple reactive groups, covering reac-tions that would occur at those groups. The mono-functional compounds that were chosen possessedvarying hydrocarbon chain lengths and degrees ofbranching for organic molecules, or different counter-ions or substituents. If conjugation or aromaticitywould be expected to influence the reactivity of agroup, such as with some highly polymerizable alde-hydes (acrolein, crotonaldehyde) and nitriles (acrylo-nitrile), a few representative unsaturated compoundswere also chosen for that functional group.

Furthermore, since 1-carbon (C1) organic mole-cules of certain functionalities (e.g., formic acid,formaldehyde, methanol, and haloforms) have a dif-ferent oxidization state on the carbon than other mol-ecules with the same functional group and less sterichindrance, their reactivity may be enhanced. Exam-ples of enhanced reactivity for one-carbon com-pounds include formic acid’s Ka being an order ofmagnitude larger than acetic acid [5] and theenhanced activity of methyl substrates in SN2 reac-tions [6]. Therefore, since the CRW is intended to pre-dict a worst-case scenario, C1 compounds were

Process Safety Progress (Vol.27, No.3) Published on behalf of the AIChE DOI 10.1002/prs September 2008 213

always selected as one of the representative com-pounds for each functional group. The increasedreactivity and common industrial use of C1 and C2compounds made them an ideal starting point forchoosing representative compounds, although com-pounds up to between C4 and C8, or up to four aro-matic nuclei for aromatics, were evaluated for eachorganic functional group, such as the three differentbutanols for alcohols, hexanes and octanes for alka-nes, and cyclohexanone for ketones.

Between 4 and 10 representative compounds werechosen for each functional group, and all reactions ofthose compounds with any one other compound,under mild conditions, (<373 K, <3 atm) and withoutthe presence of other compounds to act as catalysts(binary reactions only), were used to revise theCRW’s binary compatibility table and to build a newset of documentation for the CRW. Table1 includeslists of representative compounds for several func-tional groups.

All reaction predictions assume concentrated solu-tions, pure solids, or a significant partial pressure of agas, unless specified otherwise. Highly forcing condi-tions were excluded from the binary compatibilitypredictions because of the CRW’s focus on responseand storage compatibility, although higher tempera-

ture and pressure reactions are often listed in com-pounds’ chemical profiles. However, reactions at non-standard but relatively mild conditions were included,reflecting the wide range of ambient temperaturespresent in various locations in the United States, tobetter reflect hazard conditions in the event of a spillor fire, and to provide an additional safety margin.An example hazard statement can be seen in Figure 1,at the end of the Results section.

LITERATURE SEARCH METHODOLOGYThe documentation for CRW 2.0 was based

entirely on a survey of available literature; no newlaboratory work was conducted to assess reactivity.Most of the sources used are also commonly avail-able, or a more extensive excerpt from the source iscommonly available in secondary literature.

Once representative compounds were chosen, apreliminary research phase was conducted using twowell-known chemical engineering encyclopedias, theKirk-Othmer Encyclopedia of Chemical Technology[5] and Ullmann’s Encyclopedia of Industrial Chemis-try [7]. Examination of these articles helped determinewhether or not sufficient information would be avail-able for a ‘‘representative’’ compound, and if not,what might be a better compound to choose. The

Table 1. Example representative compounds for four reactive groups.

Inorganic Nonoxidizing Acids Aldehydes Alcohols and Polyols Active Metals

Hydrochloric acid Formaldehyde Methanol CalciumHydrobromic acid Acetaldehyde Ethanol MagnesiumHydrofluoric acid Acrolein 2-Propanol ZincPhosphoric acid Crotonaldehyde n-, sec-, and t-Butanols Iron

Benzaldehyde Ethylene glycol NickelGlycerol Cobalt

TitaniumAluminum

Figure 1. Example binary combination hazard statement.

214 September 2008 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.27, No.3)

preliminary phase also provided citations for papersthat may be useful in analyzing reactivity hazards,which were later used in the specific phase. Whenspecific reactivity information was found in the ency-clopedias, it was also recorded for use in the CRWdocumentation.

The second phase in the documentation researchconsisted of examination of general sources, includ-ing textbooks on organic (Solomon & Fryhle’s Or-ganic Chemistry [6]) and inorganic chemistry (Mellor’sModern Inorganic Chemistry [8]), and specific mono-graphs on individual functional groups or chemistryof an element, including Saul Patai’s Chemistry of theFunctional Groups series [9]. While specific documen-tation of hazardous reactions was rarely found inthese sources, the general survey phase enabled bet-ter interpretation of reactive hazards reported in othersources.

Once sufficient background information was avail-able to interpret reports of reactivity incidents, themain source used to compile the hazard documenta-tion was Bretherick’s Handbook of Reactive ChemicalHazards [10], which contains an extensive, cross-refer-enced annotated bibliography of literature about haz-ardous reactions, major accidents in the chemicalindustry, and laboratory accidents. The less-compre-hensive, but still extensive National Fire PreventionAssociation Publication 491M [11], covering hazardousreactions relevant to fire prevention, was also used asa supplemental secondary source for building thehazard statements.

Initially, two other secondary sources common inhazardous materials literature, Sax’s Dangerous Proper-ties of Industrial Materials [12], and The Sigma AldrichLibrary of Chemical Safety Data [13], were examined assources for hazard documentation, but while valuableas quick-references for safety data, neither of thosesources contained sufficient information about condi-tions under which hazardous reactions occurred norcitations of a primary source. Furthermore, hazard clas-sifications were often broad and difficult to elaborateupon, and so use of those two sources rapidly desisted,although some better-supported citations are presentin the hazard statements.

The preliminary, secondary-source search was fol-lowed by a search of primary sources if the second-ary-source information was ambiguous or insuffi-ciently complete. Abstracts of papers obtainedthrough Chemical Abstract Services were the mostcommon verification source, with full-texts of papersexamined only if a hazard was still ambiguous or aparticular paper seemed probable to have extensiveinformation about additional hazards. Other verifica-tion sources included the Manufacturing Chemists’Association (MCA) Safety Data Sheets [14] and theMCA Case Histories of Accidents in the ChemicalIndustry [15].

If verification information was unavailable or aparticular reaction was still ambiguous, CRW staffused information of physical and chemical propertiesof the compounds, from sources such as The MerckIndex [16], Hawley’s Condensed Chemical Dictionary[17], NIST WebBook [18], and The CRC Handbook of

Chemistry and Physics [19], and best available chemi-cal judgment to decide how to classify hazards.

LITERATURE SEARCH RESULTSThe representative-compound search significantly

increased the amount of available documentation forthe CRW. Of 540 binary pairs with at least one haz-ardous reaction listed, 297 (55%) now have links tocitations for at least one of the hazards, and many ofthe binary pairs are extensively documented.

Several reactive groups, including oxidizing acids,alcohols, bases, active metals, ethers, and inorganicoxidizing agents, now have a particularly extensiveamount of documentation (>80% of known interac-tions with other CRW reactive groups documentedand cited). However, documentation is very limited(<40% documented) for other important reactivegroups, including carbamates, cyanides, thiocarba-mates, isocyanates, and nitrides/phosphides/silicides/carbides, and somewhat sparse (<60%) for otherssuch as organic sulfides, peroxides, and carboxylicanhydrides, because of lack of information about spe-cific reactive hazards.

For some of the groups with less documentation,such as nitriles and peroxides, large amounts of gen-eral information about reactivity was available, butspecific information needed to assign hazard classeswas difficult to locate. For other groups, such as car-bamates, isocyanates, and organic phosphates, exten-sive toxicology, synthesis, and environmental fateinformation is readily available in the hazardousmaterials literature, but little documentation of reac-tive accidents or hazards of the use of those com-pounds in intentional chemistry could be found. Athird set of groups, including organometallics andsalts, contain such a wide variety of compounds,such that drawing general conclusions was difficult.Further work may need to be done in future to refinegroups such as organometallics, by subdividing themdue to inherent chemistry.

Another new property of the documentation is abrief description of conditions for many of the reac-tions. Representative reactions were chosen to beunder mild conditions that could be expected in astorage or spill situation, but solvent information waslisted, if relevant, for some of the reactions. The tem-peratures under which a reaction is known to occurare typically listed as ranges (see Table2), althoughspecific temperatures were used in some hazardstatements. Reactions taking place over 373 K were

Table 2. Temperature ranges for classification in CRWdocumentation.

Classification Temperature Range

Cryogenic <173 KLow 173–273 KAmbient 274–308 KSlightly above ambient 309–323 KSlightly elevated 324–343 KElevated 3441K


not used to determine binary pair hazards, nor werereactions at pressures greater than 3 atm.

While the documentation was intended to removegaps from previous versions as well as add new in-formation, nondocumented hazards from older ver-sions of the CRW were not expunged, unless compel-ling reasons for reclassification or removal werefound in the literature. This is because the CRW takesa very conservative approach to reactive hazards, anda documentation check for one of the listed, butcurrently undocumented hazards would reveal thelack of citations, and allow users to make a decisionwhether or not to disregard the hazard. An examplehazard statement appears as follows (Fig. 1).

CRW DESIGN IMPROVEMENTSReresearching the reactive groups in the CRW pro-

vided not only an opportunity to expand the product,but also to error-check and optimize the compatibilitytable, reactive group profiles, and the chemical pro-files of several compounds.

First, a number of compounds were misclassified,or questionably classified, in earlier versions, and 22compounds have now been reclassified to differentreactive groups or had additional reactive groupsadded to their profile.

Second, a number of minor errors have been cor-rected within reactive group profiles, and 143 new orupdated citations for reactions have been added toindividual compounds’ chemical profiles if the reac-tion were relevant to the compound but could not begeneralized to a reactive group. However, reactionsadded to a binary pair’s hazard statement were notadded to chemical profiles in this release, since theinformation is available in the hazard report. CRW 2.0will also have the ability to display all of the hazard-ous reactions for a reactive group in a single list, forexamining general incompatibilities instead of onlybinary pairs.

Third, the reaction hazard codes used to generatehazard reports have been streamlined. Several unusedor redundant reaction hazards have been removed,

and identified reactions merged with other hazards orassigned new hazards. One new hazard was alsoadded, and several hazards have been rewritten forclarity. The coding system was also simplified, suchthat only one code related to gas generation, violentreactions, or fire hazards would be used for any par-ticular binary pair, with lesser hazards implied by themore stringent hazard statements; e.g., if two chemi-cals are likely to ignite on contact (code B4), theyalso pose a fire hazard (code B1). Additionally, exo-thermic reactions that may be vigorous or violenthave now been flagged as such (the code for violentreactions was unused until Version 1.7), or if an ex-plosive reaction hazard is present, the appropriateexplosion code. Conversely, reactions that were al-ready defined as explosive have also been defined asexothermic if they were not already.

Finally, the CRW has been previously criticized [28]for its high rate of returning false-positives, includinghazards that seem nonsensical. These erroneousresults are due to both the limitations imposed by abinary compatibility matrix, and ‘‘worst case’’ assump-tion for CRW predictions; the CRW is designed to bevery conservative in returning a ‘‘no reaction’’ result.Few documented reactions (usually two) are neededto declare a reactive hazard between two reactivegroups, and therefore CRW 2.0 will still have a highfalse-positive rate. However, since most of the haz-ards returned by CRW 2.0 are now extensively docu-mented, the false-positives are now mitigated by theuser’s ability to view documentation and determine ifthe reactions found for the representative compoundsthat we used to determine the hazard are relevant tothe compounds the user intends to work with. Theextent of the new documentation is summarized inTables3 and4. Since the old CRW did not includeextensive documentation, the false-positive deficiencyin CRW 2.0 should be significantly lower than in anyprevious version; however, the CRW should still notbe used as the sole source for process safety data,and in a process safety use, ambiguous positiveresults should be confirmed through MSDS info, addi-tional literature searches, thermodynamic modeling,or experiment. Known issues regarding the CRW arediscussed further in the next section.

REMAINING DESIGN LIMITATIONS/DEFICIENCIES OF THE CRWWhile the CRW 2.0 update greatly expands upon

the capabilities of the already-effective CRW, severaldeficiencies that could be addressed by future ver-sions or are better covered by other products remain.Some could be addressed by incremental updates

Table 3. Documentation quality summary.

Grade No. of Reactive Groups

A (>80% documented) 5B (60–79%) 14C (41–59%) 12D (<40%) 12

Table 4. Five best and worst documented reactive groups.

Best Documented Groups % Documented Worst Documented % Documented

Ethers 90 Organometallics 19Inorganic Oxidizing Acids 88 Chlorosilanes 22Inorganic Oxidants 86 CFCs 25Alcohols 82 Phenols 25Active Metals 81 Thiocarbamates 26


using current methodology, and others are due to thesimplifying assumptions used to develop a work-ing product that can handle such a wide variety ofreactions.

First, 45% of the pairs with a reactive hazard listedin the CRW are currently undocumented, and findingother secondary-source documentation for thosereactive pairs may be difficult for users wishing toverify an undocumented CRW prediction. However,many of these reaction hazards seem intuitive and/ormatch the predictions of other compatibility tables;for example, the CRW has no documentation for areaction of strong reducing agents and amines, butthe EPA Method [3] predicts an incompatibility. Defi-ciencies for less-common reactions have generallybeen mitigated by the new documentation.

Second, predicting reaction products is currentlybeyond the scope of the CRW, although future ver-sions may include a feature for identifying gases gen-erated in a reaction. Some products, especially gases,and some explosive products, are listed in the CRW2.0 reaction documentation, but the CRW does notpredict the identity of many dangerous products gen-erated by reactions. Furthermore, such a lack ofproduct prediction contraindicates the addition of arecursive search for interaction of reaction productswith other products or reactants in a mixture,whether or not the first reaction is hazardous. Sincereactions with gaseous products can lead to particu-larly severe incidents, such as the 1995 Powell Duf-fryn fire and spill in Savannah, Georgia, where H2Sevolution was a major concern [21], a feature for pre-dicting gases generated by reactions will likely beadded to a future version of the CRW.

Third, the CRW does not currently provide quanti-tative or semiquantitative information about reactionthermodynamics or kinetics; however, users canobtain thermodynamic information about reactionswith other common safety products such as CHETAH[22]. This feature deficiency in the CRW could beaddressed by other products or by incorporating ba-sic thermodynamic prediction and/or quantitativestructure–activity relationship functionality into afuture version of the CRW.

Fourth, since the CRW functions by comparing bi-nary reactions of reactive groups, based on (usually)monofunctional representative compounds, reactionsspecific to multifunctional compounds may beneglected, as are reactions involving three or morecompounds, or presence of a catalyst, potentiallyleading to false negatives [20]. Effects of deactivatinggroups and inhibitors may also be ignored, leading tofalse positives. Nonbinary reactions are considered tobe outside the current scope of the CRW.

To utilize the CRW most effectively, users shouldread all documentation for the CRW before use, tounderstand its limitations, and when interpretingreaction predictions, consult the chemical profiles forindividual compounds or use an additional source ofreactivity data that does include catalyzed reactionsand reactions involving more than two reactants,such as Bretherick’s, NFPA 491M, the major chemicalengineering encyclopedias, or primary literature.

CONCLUSIONUse of representative compounds to draw general

conclusions about the reactivity of functional groupsenabled significant improvements to the CRW, includ-ing verification of many of the hazards previouslyidentified by the CRW, and addition of hazard state-ments for many new dangerous combinations, with-out having to re-research more than a few hundredsof the 6,0001 compounds in the database. Research-ing hazards for the selected compounds and reex-amining the database also allowed correction of a sig-nificant number of errors and omissions from earlierversions, and allowed optimization of the hazardcodes and user interface. While CRW 2.0 still has sev-eral deficiencies, the addition of hazard statementsgenerated from representative compound informationgreatly expands upon the utility of the product,which will be available as a stand-alone FileMakerdatabase in Fall of 2008.

LITERATURE CITED1. J. Farr, W. Freeman, and S. Odojewski, New pro-

gram for chemical compatibility, Chem Health Saf5 (1998), 33–36.

2. EPA 600/2–80-76, A method for determining thecompatibility of chemical mixtures, EPA, 1980.

3. CAMEO Chemicals, http://cameochemicals.noaa.gov/, National Oceanic and Atmospheric Administra-tion, 2007.

4. NOAA Chemical Reactivity Worksheet v.1.7,NOAA OR&R HMRD, 1998–2006.

5. K. Othmer, Kirk-Othmer Encyclopedia of Chemi-cal Technology, Wiley, New York, 2006.

6. T.W.G. Solomons and C.B. Fhryle, Solomons andFryhle’s Organic Chemistry, 8th ed., Wiley, NewYork, 2004.

7. Ullmann’s Encyclopedia of Industrial Chemistry,Wiley-VCH Verlag, Weinheim, Germany, 2006.

8. J.W. Mellor, Mellor’s Modern Inorganic Chemistry,6th ed., Wiley, London, U.K., 1967.

9. S. Patai (Editor), The Chemistry of FunctionalGroups (series), Wiley, Chichester, U.K., 1964–1994.

10. P.G. Urben, Bretherick’s Handbook of Reac-tive Chemical Hazards, Butterworth-Heinemann,Oxford, 1995.

11. National Fire Protection Association, NFPA Publi-cation 491M, NFPA, Boston, 1975.

12. R.J. Lewis Sr. (Editor), Sax’s Dangerous Propertiesof Industrial Materials, 8th ed., Van NostrandReinhold, New York, 1992, p. 1893.

13. R.E. Lenga (Editor), The Sigma Aldrich Library ofChemical Safety Data, 2nd ed., Sigma Aldrich,St. Louis, MO, 1988.

14. Chemical Safety Data Sheets 1–50, ManufacturingChemists’ Association, Washington, DC, 1960, 1962,1970, 1975.

15. Case Histories of Accidents in the ChemicalIndustry, Manufacturing Chemists’ Association,Washington, DC, 1962–1975.

16. S. Budavari (Editor), The Merck Index, 12th ed.,Merck, Whitehouse Station, NJ, 1996.


17. R.J. Lewis Sr. (Editor), Hawley’s Condensed ChemicalDictionary, 12th ed., Van Nostrand Reinhold,New York, 1996.

18. NIST Chemistry WebBook, http://webbook.nist.gov/chemistry/ (Standard Reference Data-base Number 59, June 2005 Release).

19. D.R. Lide (Editor), CRC Handbook of Chemistryand Physics, 80th ed., CRC Press, Boca Raton, 2000.

20. D. Quigley, F.S.H. Whyte, L. Boada-Clista, andJ.C. Laul, Use and misuse of chemical reactivity

spreadsheets, Chem Health Saf 13 ( 2006), 29–34.

21. Report HMRAD 95–9, Powell Duffryn incidentreport: Observations and chemistry, NOAA,1995.

22. C.A. Davies, et al. CHETAH–The Computer Pro-gram For Chemical Thermodynamics and EnergyRelease Evaluation, ASTM, Philadelphia, 2005,http://www.southalabama.edu/engineering/chemical/chetah/index.html.


crw 2.0: a representative-compound approach to functionality-based prediction of reactive chemical...

Documents