The Integration of ProcessSafety into a Chemical ReactionEngineering Course: KineticModeling of the T2 IncidentRonald J. Willey,a H. Scott Foglerb, and Michael B. CutlipcaDepartment of Chemical Engineering, Northeastern University, Boston, MA; r.willey@neu.edu (for correspondence)bDepartment of Chemical Engineering, University of Michigan, Ann Arbor, MI; sfogler@umich.educDepartment of Chemical Engineering, University of Connecticut, Storrs, CT; michael.cutlip@uconn.edu

Published online 6 December 2010 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/prs.10431

The explosion and subsequent death of four peopleat the T2 Laboratories, chemical facility in Jackson-ville, Florida, USA, in 2007 has resulted in the UnitedStates Chemical Safety Board finding that undergrad-uate chemical engineering students do not receiveadequate knowledge in the hazards associated withchemical processing. This article summarizes theevents that led up to the T2 tragedy. A reactor engi-neering analysis of the event is presented that can beused in a chemical reaction engineering classroom todemonstrate the hazards involved when dealingwith exothermic reactions and methods to mitigate.� 2010 American Institute of Chemical Engineers Pro-cess Saf Prog 30: 39–44, 2011

Keywords: process safety education; runaway reac-tions; reactor explosions

INTRODUCTION

Description of the AccidentOn December 19, 2007 at 1:33 PM, a tremendous

explosion shocked the northern Jacksonville, Floridaregion. Tons of sodium, hydrogen, and organicsexploded into surrounding environment via a reactorrupture and caught fire [1]. The explosion occurred ata chemical producer called T2 Laboratories. Four

people lost their life. Thirty-two people within the vi-cinity suffered injuries. The power of the shock wavewas felt 15 miles away. A video about the event isavailable to the public via the CSB website [2]. Figure1 shows the resultant plant damage a day after theevent. The plant manufactured methylcyclopenta-dienyl manganese tricarbonyl, a gasoline antiknockadditive, for a third party distributor. At the time, a2,450 gallon batch reactor was producing the first in-termediate, sodium methylcyclopentadiene andhydrogen (a by-product) from the reaction of metallicsodium with methylcyclopentadiene. Diethylene gly-col dimethyl ether (diglyme) was used as a solvent.This was the 175th batch manufactured by thefacility.

The Major Deviation that Fateful DayThe underlying direct cause of the accident was

the loss of cooling water to a reactor processing anexothermic reaction. A jacketed reactor, with the abil-ity to transfer heat generated by reaction, was used.Figure 2 shows a cross-section of the reactor. Waterwas added to a cooling jacket that surrounded the re-actor. Vaporization of the water inside the jacket car-ried away the heat released by the main reaction.Around 1:23 PM, a reactor operator contacted theone of the plant owners (a chemical engineer)informing him that there was no cooling water flowto the reactor.

What was normally a multihour exothermic pro-cess done at 3508F (176.78C) became a 10-min exo-

This was provided to us anonymously by a reviewer of the first draft ofthis work.

� 2010 American Institute of Chemical Engineers

Process Safety Progress (Vol.30, No.1) March 2011 39

thermic process. The reaction rate accelerated asdescribed by the Arrhenius effect. More complicated,and unknown to the plant employees, was a criticaltemperature, 3908F (198.68C), where a second, stron-ger, exothermic reaction kicked in, the decompositionreaction of diglyme either catalyzed or enhanced bythe presence of sodium. This rapid decompositionled to the creation of gases and vapors at a very rapidrate. Quickly, the rate of gas generation exceeded the4"-diameter relief system’s ability to relieve the reac-tor without further build up of pressure and eventualexplosion. Within minutes of the recognition of thehazardous situation, the reactor exploded.

CSB Major Recommendations IncludeChanges in Safety Education ofChemical Engineers

A major finding from the CSB investigation wasthat fundamental faults lie within the educationalpreparations of chemical engineering students. Theowners and engineers involved with T2 had littlebackground in dealing with reactive chemicals tracingall the back to their chemical engineering undergrad-uate educational preparation. The CSB recommenda-tions were [3]:

American Institute of Chemical Engineers2008-03-I-FL-1: Work with the Accreditation Board

for Engineering and Technology to add reactive haz-ard awareness to baccalaureate chemical engineeringcurricula requirements.

2008-03-I-FL-2: Inform all student members aboutthe (AIChE) Process Safety Certificate Program andencourage program participation.

Accreditation Board for Engineering and Technology2008-03-I-FL-3: Work with the American Institute of

Chemical Engineers to add reactive hazard awarenessto baccalaureate chemical engineering curricularequirements.

Present Resources Available forSafety Education

A major resource group for safety education sour-ces is through the Center for Chemical Process Safetyof the AIChE and its subcommittee called Safety andChemical Engineering Education. This group ofindustrialists, academics, and government representa-tives has been offering materials for chemical engi-neering education for more than 20 years. Their web-site is http://www.sache.org. Publications describ-ing the group’s efforts are available in Refs. 4–11, and12.

Reactor Analysis/ModelingThe analyses of exothermic reactions are not

trivial. If heat transfer is involved, independent var-iables on conversion included all reactor startingconcentrations, reactor temperature, coolant tem-perature, and time. All have nonlinear effects onconversion. For example, increasing temperaturecan accelerate a reaction rate when temperaturecontrol is lost. The heat released by the reaction isabsorbed by the contents within the reactor. Thecontents heat up. The reaction goes faster. The con-tents heat up more. The reaction goes even faster.This response is commonly known as a runawayreaction. Under severe circumstances, reactor fail-ure can occur due to high pressure as a result ofgaseous product generation and the presence ofhigh temperatures. Each species within the reactorvolume has a differential equation based on themass (mole) balance. The energy balance aroundthe reactor creates another differential equation.When the vapor phase is included, at least onemore differential equation evolves. Finally, there isthe challenge of determining properties, especiallywhen the mixture is nonideal. Sophisticated pro-grams are available that can combine propertiesprediction and exothermic reaction analysis to-gether such as SuperChems by IoMosaic [13]. Anexample of the use of SuperChems in the modeling

Figure 1. Overview of the accident scene shortly after the event. (CSB report, p. 3 Ref. 1).

40 March 2011 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.30, No.1)

of a complex exothermic reaction is presented in apaper by Willey et al. [14]. However, this programis not easily accessible to chemical engineeringstudents and requires substantial instruction tounderstand. In this work, we explore an intermedi-ate approach that offers students more accessibletools.

METHODS

Reaction Stoichoimetry SelectedThe CSB report offers details of the chemistry

about the reactions used at T2 in the report’s Appen-dix A [1]. The exothermic reaction occurring at thetime of the accident was related to the first step: the

Figure 2. Cross sectional view of the reactor. Note that the jacketed cooling zone (light grey) which used evapo-ration of water to remove the heat generated by the reaction. (CSB report, p. 20, Ref. 1).

Process Safety Progress (Vol.30, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2011 41

combination of sodium and methylcyclopentadieneto sodium methylcyclopentadiene and a 0.5 mole ofhydrogen. Further, the report noted that a second,previous unknown, exothermic reaction followed thedecomposition of the solvent, diglyme, in the pres-ence of sodium. Decompositions are very compli-cated. They do not go completely to the lowestenergy products such as water or carbon dioxide. Of-ten, a complex residue is left behind composed ofcarbon and hydrogen, and is essentially, best referredto as tar. The CSB report did not present the detailedchemical analysis of the products produced whendiglyme decomposes, however, diglyme contains 3atoms of oxygen, 6 atoms of carbon, and 14 atoms ofhydrogen. We used this to imagine that when thediglyme decomposed, three moles of permanent gas(either CO or H2) formed. It is this formation of apermanent gas, into confined space along with theincreasing vapor pressure due to higher temperaturesthat created the rapid rise in pressure, and subse-quent vessel failure.

The CSB report provided figures of reaction calo-rimetry studies completed for pure diglyme in thepresence of sodium, and for a reactant mixture repre-sentative of normal operation. A VSP2 calorimeterwas used in which background can be read in a PSParticle by Theis et al. [15]. The VSP2 results demon-strated conclusively that the diglyme in the presenceof sodium rapidly decomposes when the temperatureexceeded 3908F. The pressure and temperature riseduring this exothermic reaction at it peak conditionswas about 32,000 psig per minute and 2,3408F perminute, respectively [1]. We modeled the second reac-tion as a first-order decomposition with an activationenergy of 80 kJ mol21. Essentially, the first trace ofexothermic activity is shown around 3588F and thestart of the runaway is noted at 3908F. For the reac-tion of sodium with methylcyclopentadiene, we esti-mated an activation energy of 40 kJ mol21, againusing VSP2 data for guidance.

Several other parameters had to be estimated toreach a working model for the two reactionsinvolved. These were the composition loaded intothe reactor (we used the same mass fractions as givenin Table 1). For the amount loaded into the reactor,we selected a level of two-thirds full based on thevideo presented by the CSB on the accident thatshows this level in the reactor [2]. This equates to6,000 dm3 (or liters) of liquid and 3,000 dm3 of ‘‘headspace’’ or gas space within the reactor. The heat of

reactions were initially estimated via Aspen simula-tion software [16], and later slightly adjusted to matchthe temperature rise shown in the CSB report [1].These values were 245 kJ per mol of sodium reactedto products and 2390 kJ per mol of diglyme decom-posed, respectively. We also needed heat capacity ofthe liquid reaction mixture plus the reactor shell thatwas estimated to be a sum of 16 MJ K21 for the entiremass of the reactor mixture and reactor mass (Smicpiof liquid contents and reactor material). As weworked on the model, we included an energy termfor the vaporization of diglyme. Vaporization tempersthe heating effect within the reactor because a por-tion of the heat released by reaction is used to vapor-ize the solvents within the reaction. For example, ifthe reactor’s rupture disk had open at 75 psig, thenthe rapid vaporization of the reactor contents wouldhave cooled the reactor and prevented the reactorcontents from moving into the critical region wherediglyme decomposition accelerates.

Reaction Model Development

Kinetic Models Used

Reaction 1 (normal process reaction).

C6H8 þ Na ! ½C6H�7 � � �Naþ� þ 0:5H2 þ heat

Kinetic rate model used: 2rC6H85 k1[CNa][C6H8],

mols dm23 min21. With k1 5 k1A exp(2E1/RT), dm3

mol21 min21

k1A5 400 dm3 mol21 min21 and E1 5 40,000 Jmol21 K21

Reaction 2 (diglyme decomposition reaction).

C6H14O3 ! 3 moles of gasþ a liquid or solid residual

þ heat

Kinetic rate model used 2rC6H14O35 k2[C6H14O3],

mols dm23 min21. With k25 k2S exp(2E2/RT) min21

k2S 5 1 3 106 min21 and E2 5 80,000 J mol21 K

Reactor TypeThe reactor type used was a batch reactor with the

ability to bleed off pressure via a 1@ line. Should thepressure exceed the rupture disk rating, the rupturedisk would burst and vent the reactor. The venting

Table 1. Batch reactor feed contents taken for the ‘‘normal’’ reactor run (VSP2 Run 6 CSB report) [1].

ChemicalVSP2 Run(grams) Weight %

Mass usedfor simulation

discussed below(kg)

Concentrationused for simulation

(mol/dm3)

Sodium 5.147 10.59 593 4.3MCDP dimer 21.164 43.55 2,452 2.55Diglyme 20.690 42.58 2,415 3Mineral oil 1.591 3.28 185 0.088

42 March 2011 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.30, No.1)

equation was only focused on the 4@ line. The ventingwas modeled proportional to the absolute pressure inthe reactor assuming critical flow of vapor through line.

Process Simulation by PolymathPolymath is a software program developed by Mi-

chael B. Cutlip et al. [17] for general use in chemicalengineering and includes differential equations solv-ing. It is geared to the chemical engineering student,and its programming is straightforward. Fogler inte-grates Polymath examples throughout his reactionchemical engineering textbook [18]. Many examplesare presented to assist students and others in the useof this important mathematical tool. Educational andprofessional versions of Polymath are available atmoderate prices [19]. Inexpensive site licenses of theeducational student version are available via theCACHE web site [20].

RESULTSThe Polymath code used in the simulation is avail-

able from the authors. Verification of the model run-ning adiabatically is compared against the CSB’s

reported VSP2 data, which is shown in Figure 3.Three scenarios were then simulated: (1) normaloperation, with a cooling water added to maintainthe reactor temperature at 3508F; (2) operation with-out cooling water (an adiabatic reactor) with a rup-ture disk opening set at 75 psig; and (3) an adiabaticreactor with a rupture disk opening set at 400 psig.Temperature and pressure profiles are shown in Fig-ures 4 and 5 for the three scenarios. It is clear thatthe reactor never exceeds 3908F if cooling water ispresent. If the rupture disk opens at 75 psig, the reac-tor continues to heat and pressurize but at a slowrate, and eventually cooling occurs due to internalvaporization. On the other hand, venting the reactorat 400 psig leads to a runaway, as the reaction pres-sure generation rate far exceeds the cooling offeredby the vaporization. The results happen very rapidly,within a few minutes once the temperature passes5008F. Figure 6 shows the conversion profiles of so-dium, which represents the normal desired reactionfor the three scenarios. Figure 7 shows the conver-sion of diglyme for the three scenarios. When run-

Figure 3. Matching of reaction model with VSP2 dataoffered for mixture as shown in the CSB T2 report,Figure 18.

Figure 4. Temperature as a function of time for thethree scenarios investigated in this work.

Figure 5. Pressure as a function of time for the threescenarios. Note the rapid pressure rise for the reliefopening occurring at 400 psig.

Figure 6. Conversion of sodium the limiting reactantwithin the reactor.

Process Safety Progress (Vol.30, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2011 43

away occurs, it is the rapid conversion of diglymethat causes the explosion.

DISCUSSIONThis case study is an excellent example of run-

away batch reactions, and can be explained to thestudents of chemical engineering at several levels. Atthe undergraduate level, one can point out that everyexothermic batch reaction deserves respect andunderstanding. That test mixtures should be run atthe small scale in thermochemistry instruments suchas Differential Scanning Calorimeter (DSC), Vent SizingPackage (VSP), and Accelerating Rate Calorimeter (ARC).That the test runs should include all reactants, solvents,and products to insure that the thermochemistry isunderstood at the microscale first. This point can bementioned even in organic chemistry classes discussingexothermic organic reactions. At the next level, under-graduate reactor engineering, the hazards associated withscale up and processing reactions in batch reactors canbe pointed out using this example. The CSB onlinevideo adds to the classroom presentation. At the gradu-ate level, students can review the details of the modelpresented in this article, look at the assumptions, andimprove on the predictability, in an effort to understandwhat details are needed to predict runaway reactions.For the practitioner, we offer techniques to simplify acomplicated system, methods to estimate activationenergy and system properties, plus the use of an easy touse differential equation solver that makes examining thesystem facile.

LITERATURE CITED1. U.S. Chemical Safety and Hazard Investigation

Board Investigation, Report No. 2008–3-I-FL Sep-tember 2009, Available at: http://www.csb.gov/investigations/detail.aspx?SID58&Type52&pg51&F_InvestigationId58? Accessed on July 5, 2010.

2. CSB, Available at: http://www.csb.gov/videoroom/detail.aspx?VID532. Accessed on July 5,2010.

3. CSB, Available at: http://www.csb.gov/recommendations/details.aspx?SID58. Accessed onJuly 5, 2010.

4. D.C. Hendershot, A history of process safety andloss prevention in the American Institute ofChemical Engineers, PSP 28 (2009), 105–113.

5. D.A. Crowl, A century of process safety—Andmore, PSP 28 (2009), 103–104.

6. J.F. Louvar, Safety and chemical engineering educa-tion—History and results, PSP 28 (2009), 131–134.

7. D.C. Hendershot and W. Smades, Safety culturebegins in the classroom, PSP 26 (2007), 83–84.

8. J.F. Louvar and D.C. Hendershot, Education mate-rials for universities and industry, PSP 26 (2007),85–89.

9. J.F. Louvar and F.O. Kubias, Safety and chemicalengineering education, PSP 13 (1994), 72–74.

10. L. Osborn, Process safety in education, PSP 18(1999), W5–W6.

11. R.J. Willey, SACHE case histories and trainingmodules, PSP 18 (1999), 195–200.

12. F.O. Kubias and J.F. Louvar, Upgrading educationand safety performance: SACHE at 17, PSP 21(2002), D2–D4.

13. IoMosaic, Available at: http://www.ioiq.com/superchems/overview.aspx. Accessed on July10, 2010.

14. R.J. Willey, F. Rodrigues, S. Chippett, G. Melhem,and S.K. Singh, Thermo-kinetic analysis of reac-tions involved in the manufacture of o-nitroani-line, PSP 20 (2001), 123–129.

15. A.E. Theis, J.P. Burelback, and C.F. Askonas,Safety scale-up processes and accommodate rec-ipe changes, PSP 28 (2009), 135–140.

16. Aspen Technology Inc. Available at: http://www.aspentech.com/. Accessed on July 10, 2010.

17. M. Shacham, Michael B. Cutlip, and M. Elly, Poly-math Numerical Computation Package, 2010.

18. H.S. Fogler, Elements of Chemical Reaction Engi-neering, 4th Edition, Upper Saddle River, NJ:Prentice Hall Inc., 1999.

19. http://www.polymath-software.com/order/index.htm. Accessed on July 8, 2010.

20. http://www.che.utexas.edu/cache/polymath.html. Accessed on July 8, 2010.

Figure 7. Conversion of diglyme for the three scenar-ios. Runaway occurs for rupture disk setting of 400psi.

44 March 2011 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.30, No.1)