Influence of trace impurities on chemical reaction hazards

Journal of Loss Prevention in the Process Industries 15 (2002) 37–48 www.elsevier.com/locate/jlp

Influence of trace impurities on chemical reaction hazards J.L. Gustin

*

RHODIA-RHODITECH 24, avenue Jean-Jaure`s F69153, Decines Charpieu, France

Abstract The influence of trace impurities is frequently mentioned as a possible or probable cause of accidents in the chemical industry. In process conditions where there is a potential for a fast exothermic decomposition or polymerisation reaction, the contamination of pure chemicals by trace impurities may cause problems. Typical examples of this situation are described concerning the processing of organic nitrocompounds and the storage of reactive monomers, i.e. vinyl acetate and ethylene oxide. In some process instances, i.e. hydrogenation of organic nitrocompounds, trace impurities in the organic substrate are said to have killed the catalyst and caused severe accidents. Many contaminants are known to be catalyst killers in hydrogenation processes. Examples of accidents or runaway reaction hazards caused by trace impurities are given, based on a review of the literature and on our own data. Trace impurities present in common raw materials make their processing more complex and may cause accidents. Two examples are considered, the presence of nitrogen trichloride in chlorine manufacture and the presence of hydrogen sulphide and mercaptans in petrochemical feed-stock. A full literature review is given.  2001 Elsevier Science Ltd. All rights reserved. Keywords: Organic nitrocompounds; Vinyl acetate; Ethylene oxide; Nitrogen trichloride; Hydrogen sulphide; Mercaptans

1. Introduction The influence of trace impurities on chemical reaction hazards is not frequently considered as a general subject to draw useful conclusions improving process safety. However, the influence of trace impurities is often discussed in chemical accident enquiries to explain the occurrence of unexpected or unwanted fast chemical reactions or decompositions. The contamination of pure chemicals by trace impurities may lower their thermal stability to a large extent and cause unexpected decomposition under normal process conditions. Representative examples of this phenomenon which have caused frequent accidents are discussed in this paper. These examples have been mentioned in our recent papers on different process safety issues. The examples considered here are the influence of trace impurities on the thermal stability of organic nitrocompounds (Gustin 1997a, 1998) and the influence of trace impurities on the thermal stability of reactive monomers with consideration of two monomers, vinyl acetate (Gustin & Laganier, 1998a,b) and ethylene oxide (Gustin, 2000). * Tel.: +33-4-72-93-57-14; fax: +33-4-72-93-53-50.

Trace impurities in raw materials may impact large chemical processes where this problem is taken into account. It is also the cause of well known or frequent accidents. Two such examples are considered in this paper, i.e. the presence of nitrogen trichloride in chlorine manufacturing processes and the presence of hydrogen sulphide and mercaptans in petroleum gas and crude oil processing.

2. Chemical reaction hazard due to contaminants in processing organic nitrocompounds Pure organic nitrocompounds, i.e. aromatic or aliphatic nitrocompounds, decompose at high temperature exhibiting a large decomposition exotherm. In most cases, the decomposition is violent or explosive. In practical process situations nitrocompounds are mixed with other chemicals or contaminated by impurities which lower their thermal stability. Contaminated nitrocompounds or solutions of nitrocompounds may decompose at much lower temperature than the pure product. Their decomposition is less rapid but remains highly exothermic. Therefore, for practical reasons, the most relevant information in the field of Process Safety is to describe

0950-4230/02/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved. PII: S 0 9 5 0 - 4 2 3 0 ( 0 0 ) 0 0 0 4 3 - 7

38

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

how contaminants or impurities may affect the organic nitrocompounds’ thermal stability and in which process situation this may be a hazard. 2.1. Pure organic nitrocompounds Pure organic nitrocompounds, aromatic or aliphatic nitrocompounds, decompose at high temperature. Their decomposition is rapid and highly exothermic. The decomposition reaction of nitrocompounds may be autocatalytic, i.e. exhibit a chemical acceleration under constant temperature conditions or follow an Arrhenius n-order rate equation. Some nitrocompounds are known to exhibit an autocatalytic decomposition. Their decomposition exhibits an isothermal induction period phenomenon at temperature just below the decomposition onset temperature under temperature scan conditions (Gustin, 1995). This can be shown using a DTA apparatus. A list of organic nitrocompounds exhibiting an autocatalytic decomposition phenomenon is given by Grewer (1994) (p. 210). When Process Safety is considered; the isothermal exposure test temperature and duration must be representative of the process conditions. Therefore, test duration may be of several weeks. Monitoring of pressure or gas generation in the test vessel is recommended because the detection of gas generation is more sensitive than the detection of heat production by the sample. A good example of the above-mentioned behaviour is the experimental investigation of a dinitrotoluene pipeline explosion by Bateman, Small & Snyder (1974). Dinitrotoluene was contained in a transfer line for 10 days and was probably exposed to temperatures of about 210°C. The transfer line finally exploded in several places. Isothermal exposure tests were performed at temperatures between 150°C and 230°C in a closed vessel. The isothermal induction period for the decomposition gas production was found to be 31 days at 150°C. This induction period showed linear variation in log scale, as a function of the temperature in reciprocal scale, thus proving the autocatalytic nature of the phenomenon. Traces of Na2CO3 are shown to lower the thermal stability of dinitrotoluene. 2.2. Thermal stability of organic nitrocompounds in the presence of other chemicals 2.2.1. In nitration process Organic nitrocompounds are obtained in nitration processes where nitric acid is reacted with organic reactants in semi-batch or continuous operations. Some nitrations are made in concentrated nitric acid. However, in most industrial processes, the organic reactant is dispersed in sulphuric acid and nitric acid or mixed acid is injected slowly in the reaction mixture as controlling reactant. The contamination of nitrocompounds with concen-

trated sulphuric acid lowers their decomposition onset temperature to a large extent, depending on the ratio of sulphuric acid to nitrocompound. An example of this effect is shown for a nitrochloro methyl ester aromatic derivative using DTA tests under temperature scan conditions on Fig. 1. An example of the influence of sulphuric acid and nitric acid on the thermal stability of nitrocompounds is the following. A 4.5-m length of 38-mm nominal bore stainless steel pipe was completely destroyed by the thermal explosion of crude dinitrotoluene. During nitration, dinitrotoluene extracts nitric acid. Some crude dinitrotoluene was enclosed in a pipe, at a temperature of about 130°C. The explosion occurred after 3.5 h, although pure dinitrotoluene is stable up to 250°C. The nitric acid concentration and the presence of free sulphuric acid were sufficient to depress the thermal stability of dinitrotoluene. leading to an explosion (Anon., 1989). 2.2.2. Washing of nitration products After the nitration step, organic nitrocompounds are separated from the acid phase and washed to remove the acids and water soluble impurities. This operation must be performed carefully to obtain uncontaminated stable nitrocompounds. If the organic phase is processed in further operations such as drying by water evaporation, solvent removal by evaporation, purification or isomeric distillation, separation of heavies or detarring, the washing and decantation process becomes critical due to the possible decomposition of contaminated nitrocompounds under high temperature conditions. Maloperation in the washing/decantation steps may cause contamination of the nitrocompounds with caustic soda and salts such as sodium sulphate, sodium nitrate, other metal salts, nitrophenols, nitrophenates, which may decrease the wanted nitrocompound thermal stability. Other impurities such as metal chlorides (NaCl, CaCl2) may be introduced by brine intrusion if brine is used as a cooling fluid. Drying may concentrate inorganic impurities in the organic phase and lower its thermal stability. Dehydrochlorination of chloro nitrocompounds under medium or high temperature may contribute to metal chloride contamination of the organic phase. The formation of heavies, polynitroderivatives, nitropolymers, oxidation products, which are not removed by the washing and decantation process may also lower organic nitrocompound thermal stability. The influence of possible contaminants on the organic nitrocompounds’ thermal stability is as follows. 2.2.3. Caustic soda Contamination of nitrocompounds with caustic soda reduces their thermal stability. As an example, Fig. 2 shows the influence of various amounts of caustic soda

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

39

Fig. 1. DTA thermograms under 5°C C/min temperature scan conditions of nitrochloro methylester aromatic derivative with different proportions of 95% sulphuric acid: 1, 4.6% wt.; 2, 10% wt.; 3, 20% wt.; 4, 27% wt. Obtained in Setaram DSC111–stainless steel closed cell. Heat flux refers to unit sample mass.

Fig. 2. DTA thermograms under 5°C/min temperature scan conditions of nitrocumene isomer with different proportions of pure caustic soda: 1, 5% wt.; 2, 10% wt.; 3, 15% wt. Obtained in Setaram DSC111–stainless steel closed cell. Heat flux refers to unit sample mass.

on the thermal stability of a nitrocumene isomer mixture, measured in DTA under temperature scan conditions. 2.2.4. Sodium sulphate Sodium sulphate has limited influence on the thermal stability of organic nitrocompounds. 2.2.5. Metallic nitrates Sodium nitrate and nitrates of other metal impurities may reduce thermal stability by initiating nitric oxidation of organics above 140°C.

2.2.6. Metallic chlorides The influence of metal ions on nitrocompounds’ thermal stability is often investigated using metal chlorides (Grewer & Klais, 1988; Grewer & Rogers, 1993) for practical reasons. Metal chlorides are water soluble and are easily dispersed in organic liquids. As shown in Grewer & Klais (1988) and Grewer & Rogers (1993), metal chlorides influence to a large extent the organic nitrocompounds’ thermal stability. As an example, Fig. 3 shows the influence of calcium chloride on the thermal stability of a nitro chloro methyl ester

40

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

Fig. 3. DTA thermograms under 5°C/min temperature scan conditions of nitrochloro methylester aromatic derivative with different proportions of CaCl2: 1, pure; 2, 1.49% wt.; 3, 4.84% wt.; 4, 13% wt. Obtained in Setaram DSC111–stainless steel closed cell. Heat flux refers to unit sample mass.

aromatic derivative, studied in DTA under temperature scan conditions. The screening of thermal stability of contaminated organic nitrocompounds using DTA in temperature scan mode does not imply that the thermal decomposition obtained follows Arrhenius n-order kinetics. In some instances, the decomposition may be autocatalytic and could be initiated at lower temperature after an isothermal induction period. 2.2.7. Formation and separation of heavies In refining steps, the formation and separation of heavies of nitro organic compounds may be a hazard. Unless the heavies are well identified products and remain liquid, in most instances the heavies are contaminated nitrated tars and solids because the process is not clean. The processing of these heavies at high temperature is a hazard. The heavies should not be allowed to accumulate and a process hazard review should be prompted on the nitration and washing steps to reduce or suppress heavies production. Accidental decomposition of heavies is frequently reported (Anon., 1994). 2.2.8. Hydrogenation of organic nitrocompounds Organic nitrocompounds are hydrogenated to produce amines. In hydrogenation processes organic nitrocompounds react with hydrogenation catalysts, carbon supported precious metals or Raney nickel, under hydrogen pressure ranging from a few bars to a few hundred bars. Contamination of nitrocompounds with hydrogen containing catalyst may cause accidents because the heat of hydrogenation by the active catalyst, 525 kJ/mol, is very large (David, 1965; Macnab, 1981; Stoessel, 1989).

In addition, as hydrogenation proceeds in normal process conditions, unstable intermediates are formed and can accumulate, which are phenyl nitroso and phenyl hydroxylamine derivatives if precious metal catalysts are used, azo and azoxy derivatives if Raney Nickel is used (Macnab, 1981). Tong, Seagrave & Wiederhorn (1977) described an incident during the hydrogenation of 3,4-dichloronitrobenzene (3,4 DCNB) in an agitated autoclave. During this operation the unstable intermediate 3,4-dichlorophenylhydroxylamine (DCPHA) could accumulate due to the reactant contamination by nitrate ion impurities. The subsequent decomposition of the accumulated intermediate caused the reactor explosion. The intermediate DCPHA mentioned in Tong et al. (1977) implies that the catalyst used was a precious metal which would be sensitive to reactant impurities. This accident was further discussed by Macnab (1981), Stoessel (1989) and Grewer (1994) (pp. 310–312). The main causes of significant unstable intermediate accumulation in hydrogenation processes are a wrong choice of catalyst or the catalyst killed by impurities, too low a concentration of catalyst, too low a hydrogen feed rate or hydrogen feed interruption, not enough agitation. The most active impurities are metal salts, metal nitrates, sulphur containing compounds, wrong lubrication oil, carbon monoxide in hydrogen. The formation of azo and azoxy derivatives may cause an increase of the reaction mixture viscosity lowering the reactor cooling capacity. When the hydrogenation reaction is fast, the probability of unstable intermediate accumulation is low. If the hydrogenation reaction exhibits an unexpected slow rate, the unstable intermediate accumulation hazard is

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

high and one should immediately investigate the causes of the abnormal situation. Fig. 4 shows the DTA thermogram of hydrogenation reaction mixtures of a nitrocumene isomer mixture. In this example, only limited unstable intermediate accumulation is obtained due to good hydrogenation conditions. 2.2.9. Decomposition of nitrocompounds in the presence of metal halides As discussed above, metal halides lower the thermal stability of aromatic nitrocompounds. AlCl3 is used in Friedel–Crafts reactions in conjunction with nitrobenzene as solvent. Nitrobenzene is also a good solvent to clean AlCl3 deposits. Several accidents occurred due to the decomposition of AlCl3-nitrobenzene mixtures (Gickel, 1983; Riethmann et al., 1976). 2.2.10. Recovery of nitrocompounds from solutions The batch concentration of nitrocompound+solvent solutions may be a critical operation, specially in second crop recovery. Several examples of explosive decomposition are known during concentration of impure nitrocompound solutions. Impurities may lower the solution thermal stability. A further unfavourable circumstance seems to be the decreasing level of the liquid solution in batch evaporation. Projection of solution on the vessel wall by the agitator may allow concentrated nitrocompounds to decompose on the hot unwetted reactor wall. The presence of active carbon used to decolorize impure product may also contribute to the decomposition initiation. Active carbon or graphite lowers the thermal stability of organic nitrocompounds. The concentration

41

of impurities on active carbon may also contribute to this effect.

3. Influence of trace impurities and construction materials on the thermal stability of vinyl acetate monomer 3.1. Runaway reaction hazard in storage vessels The thermal stability of reactive monomers in bulk storage vessels is very dependent on the presence of a polymerisation inhibitor and of trace impurities which can speed up the inhibitor consumption. In some process circumstances a polymerisation initiator may also be present. Any monomer is particular and should be considered separately. However the example of vinyl acetate is representative of the safety issues in this field. Vinyl acetate (VAM) is processed to produce polymers and copolymers used in water based paints, adhesives, paper coatings or non-woven binders and various applications at moderate temperatures. The polymerisation processes used include semi-batch or continuous, solution, suspension and emulsion processes. The bulk polymerisation of vinyl acetate is not used in industrial operations due to the violent reaction obtained. However the unwanted bulk polymerisation of vinyl acetate is involved in many industrial accidents where the polymerisation occurred in storage vessels containing fresh or recycled monomers or in premix vessels where a polymerisation initiator was dissolved in vinyl

Fig. 4. DTA thermograms under 5°C/min temperature scan conditions of hydrogenation reaction mixtures of a nitro cumene isomer mixture. Thermal stability as a function of residual nitro derivative concentration. 1, 86.5% wt.; 2, 56% wt.; 3, 25.5% wt.; 4, ⬍100 ppm nitrocompound. Note the low unstable intermediate accumulation. Stainless steel closed cell. Heat flux refers to unit sample mass.

42

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

acetate monomer, previous to its use in the polymerisation process. The bulk polymerisation of vinyl acetate in storage vessels occurs spontaneously under constant temperature conditions, due to a chemical acceleration phenomenon related to the free radical nature of vinyl acetate chain polymerisation. 3.2. Thermal stability of commercial monomers Vinyl acetate is a reactive monomer which may undergo a free radical chain polymerisation phenomenon under constant temperature conditions. The chain polymerisation of monomers may be initiated by radical initiators, i.e. inorganic or organic peroxides, azobisisobutyronitrile (AIBN), others % In monomer storage vessels, there is no such initiator in normal process conditions. However, there is a slow thermal production of free radicals in the bulk liquid monomer even under ambient temperature. The thermally produced radicals may accumulate and further initiate the free radical chain polymerisation of the monomers in the storage vessel. To provide enough thermal stability, some polymerisation inhibitor is added to the commercial product, which is a radical scavenger. The polymerisation inhibitor used for the stabilisation of vinyl acetate monomers is hydroquinone (HQ). RHODIA is a leading supplier for hydroquinone and other polymerisation inhibitors. Two grades of vinyl acetate monomers are currently supplied: 앫 the low hydroquinone grade containing 3–7 ppm HQ to be used within 2 months of delivery 앫 the high hydroquinone grade containing 12–17 ppm HQ for storage up to 4 months before use. Compared with other vinylic or acrylic monomers, i.e. acrylic acid (AA), methacrylic acid (MAA) or methyl methacrylate (MMA), the polymerisation inhibitor concentration in VAM is very low owing to the low thermal activity of VAM in storage conditions. 3.3. Influence of oxygen Many questions arise concerning the recommended storage conditions of vinyl acetate. It is sometimes recommended to store vinyl acetate monomer under ambient air atmosphere to enhance the polymerisation inhibitor efficiency as it is known that quinonic inhibitors need oxygen to be active. For example, air is bubbled in acrylic acid storage vessels to allow the polymerisation inhibitor, hydroquinone mono methyl ether (MeHQ) to be active (Levy, 1987). However, contrary to acrylic acid which is stored at a temperature below 25°C and has a flash point of 49°C

(INRS, 1980) above the storage temperature, vinyl acetate monomer is a highly flammable liquid with a flash point of ⫺7°C (INRS, 1980). Under ambient temperature conditions, the storage vessel gas phase under air atmosphere would be flammable for VAM. Therefore, storage of VAM under air atmosphere should be avoided. It was further pointed out by Nicholson (1991) that the optimum stability of methacrylic acid in the presence of HQ or MeHQ as polymensation inhibitor was obtained at much lower equilibrium oxygen concentration than that provided by an air atmosphere. The most favourable oxygen concentration was shown to depend on the inhibitor considered and on its concentration in the monomer. A higher oxygen concentration than the optimum value was shown to decrease the monomer thermal stability due to unstable peroxide formation. Therefore, any monomer/inhibitor couple is specific and requires a special consideration for the choice of storage conditions. In the case of vinyl acetate containing hydroquinone as an inhibitor, it was later pointed out by Levy (1993) and Levy & Hinojosa (1992) that the optimum thermal stability was obtained under dry nitrogen atmosphere without any oxygen present in the storage vessel gas phase. It was shown that between 30°C and 120°C, the length of vinyl acetate polymerisation isothermal induction period is about ⫺0.4 order in oxygen partial pressure. The detrimental effect of the presence of oxygen in the monomer gas phase, on the thermal stability of vinyl acetate containing hydroquinone as an inhibitor, is explained by the low thermal stability of the peroxides formed when VAM is stored in the presence of dissolved oxygen, The decomposition of accumulated peroxides near ambient temperature can induce an increased inhibitor consumption. Therefore, for flammability and thermal stability reasons, it is recommended to store vinyl acetate under dry nitrogen blanket. The above conclusion holds for high temperature (100°C) and ambient temperature conditions. However, when stabilised with 3–5 ppm HQ, both air saturated and oxygen-free VAM exhibit adequate thermal stability at normal transport and storage temperature (25–50°C).

3.4. Influence of impurities

It was shown by Levy that acetaldehyde impurity does not cause oxygen induced destabilisation of vinyl acetate (Levy & Hinojosa, 1992). The presence of humidity in vinyl acetate would produce some degree of hydrolysis and alter the monomer quality.

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

Table 1 Vinyl acetate in a 2-l coloured glass vessel Time (days)

HQ concentration (mg/kg)

0 5 18 41 55 67

20.1 19 18.4 17.8 18 17.6

3.5. Influence of the storage vessel wall material As for other reactive monomers and especially in the case of polymerisation accident inquiries (Kurland & Bryant, 1987) the influence of the storage vessels and transport containers wall material has been investigated. The influence of carbon steel on the thermal stability of VAM containing 5 ppm HQ, at 48.9°C, blanketed with 10% oxygen, was investigated by Levy & Hinojosa (1992). It was found that untreated carbon steel covered with rust has a strong destabilising influence on VAM compared with a glass vessel, whereas clean carbon steel would increase thermal stability. The influence of carbon steel on vinyl acetate thermal stability can also be measured by monitoring of the inhibitor concentration as a function of time. Two 2-l samples of vinyl acetate containing approximately 20 ppm hydroquinone were submitted to aging at 20°C, one in a coloured glass vessel the other in a 2-l oxidised carbon steel vessel. Both vessels were closed under air atmosphere. The depletion of HQ concentration, measured by UV spectrometric determinations, is shown in Tables 1 and 2 for the coloured glass vessel and the carbon steel vessel, respectively. In these experiments the hydroquinone consumption in the carbon steel vessel was twice that in the coloured glass vessel. Also, enough inhibitor concentration would be present under ambient temperature to prevent polymerisation over a very long period of time in both containers. However, considering the lower thermal stability of VAM in oxidised carbon steel vessels, it is reTable 2 Vinyl acetate in a 2-l oxidised carbon steel vessel Time (days)

HQ concentration (mg/kg)

0 1 7 12 22 48 61 76

21.7 20.8 19.2 19.1 17.6 16.5 16.5 15.9

43

commended to design stainless steel storage vessels for VAM and to protect existing carbon steel storage vessels from rust. This should be easier under a dry nitrogen blanket. 3.6. Accidental polymerisation of recovered vinyl acetate The data found in the literature and additional data provided, show that commercial vinyl acetate should never polymerise in normal storage conditions. However, polymerisation accidents in storage vessels are known concerning VAM which is not exactly the commercial product. This is the case for vinyl acetate recovered from polymerisation processes where the conversion ratio is not 100%. An example of such a process is the manufacture of polyvinyl alcohol by polymerisation of vinyl acetate in the presence of methanol initiated by azobisisobutyronitrile (AIBN), followed by alkaline hydrolysis of the polyvinyl ester. The unreacted vinyl acetate is separated from the polymer and recycled to the polymerisation step. The recovered vinyl acetate is free of polymerisation inhibitor and may possibly contain some traces of polymerisation initiator. A 35-m3 storage container of recovered VAM polymerised during the summer vacations in a southern country. The long residence time and the warm temperature, possibly 1 month and 50°C, allowed the polymerisation to occur. The polymerisation was violent, The storage vessel roof was ejected to 100 m and polymers were spread in the neighbourhood. The conditions where this accident occurred are fairly compatible with the stability data of Levy for VAM with no HQ, 10–20% vol. O2 and the presence of carbon steel wire. Also it was shown that oxidised iron lowered the product thermal stability measured in DTA stainless steel closed cell under 5°C/min temperature scan. The polymerisation exotherm onset temperature was shifted from 300°C to 230°C in the presence of oxidised carbon steel. The product polymerisation exotherm was 1221 J/g, close to the literature data of 1036 J/g (Encyclopedia of Polymer Science and Engineering, 1988). A sample of recovered VAM showed polymerisation isothermal induction periods at high temperature in DTA (see Fig. 5) but such data cannot be extrapolated to ambient temperature where induction periods are too long to be measured using DTA machines. Our comment on this incident is that one should take care of the thermal stability of non commercial or recovered vinyl acetate, in particular if high storage temperature and long residence time are possible. The storage of uninhibited monomer should be avoided. 3.7. Polymerisation incidents in premix vessels It is a surprising practice, from a process safety point of view, that concentrated solutions of polymerisation

44

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

Fig. 5. Recovered VAM—polymerisation isothermal induction period measured in DTA stainless steel closed cell. Curve 1, 300°C; curve 2, 290°C; curve 3, 280°C.

initiators are used in premix to be injected in a continuous or semi-batch polymerisation process. Premix vessel polymerisation accidents are known for many reactive monomers. This type of incident has occurred also in vinyl acetate polymerisation processes. 1.8% wt. dilauroylperoxide was dissolved in vinyl acetate in a premix to be injected in a VAM polymerisation process. The premix was prepared in advance, to be used 8 h later. Due to the hot summer temperature, the premix polymerised. The reaction was extremely violent. A two-phase flow of polymer and vapour was ejected in the vessel vent header, The release did not ignite. This incident has been investigated from an experimental and a theoretical point of view to assess the conditions of occurrence and to evaluate the venting requirement of this scenario. Our conclusion on this issue is that the practice of dissolving the polymerisation initiator in the reactive monomer in a premix should be eliminated from polymerisation processes for process safety reasons.

4. Influence of trace impurities and construction materials on the thermal stability of ethylene oxide Ethylene oxide (EO) is a reactive monomer obtained by oxidation of ethylene. In ethoxylation processes, EO is reacted with various chemicals to produce surfactants, solvents and specialty chemicals, Due to the thermal instability of EO in the liquid phase and in the gas phase and to the high reactivity of the monomer with other chemicals including water, many accidents have

occurred in both the manufacture and the processing of ethylene oxide. 4.1. Accidents occurred in EO production plants It is of interest to review accidents occurred in EO production plants because they may have common causes with accidents occurred in ethoxylation plants. A well-known ethylene oxide plant accident is the explosion at Doe Run Olin Mathieson plant on 17 April 1962 (Troyan & Levine, 1968). In this accident, a 26-m3 EO storage tank ruptured violently. The domino effects involved a vapour cloud explosion and severe mechanical damages to the neighbouring installations. The cause of the EO violent reaction is a back-flow of aqueous ammonia from an ethanolamine tube reactor to the EO storage tank. As EO reacts violently in the presence of small amounts of ammonia, it is necessary to install reliable back-flow prevention from the ethoxylation reactor to the EO feed tank. Also the EO feed is best located in the reactor gas phase. Any back-flow of ethoxylation reaction mixture to an EO storage tank would induce a similar runaway reaction hazard. 4.1.1. Explosion at BASF Antwerp (Basant) EG/glycol plant (Anon., 1991) A severe explosion occurred at the Basant plant on 7 March 1989 causing extensive destruction. The reconstruction of the accident circumstances proved that the explosion originated in an insulated level indicator of the aldehyde column. An EO leak in a wet insulating material with rust deposits produced polyethylene gly-

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

cols at temperatures of 80-120°C. It was shown later that PEG with a chain length greater than four monomers can oxidise in glass wool and produce a hot spot above 600°C—a temperature high enough to initiate the decomposition of EO in vessels and pipes (Schliephake, Giesbrecht & Loffler, 1992). The lessons learned from the Basant explosion were to use high integrity piping and vessels of welded construction in EO plants, to check carefully for leaks after maintenance, to avoid stagnant liquid in small pipes if temperature is above 80°C to choose carefully insulation material, to check that pipes and insulation are rust-free, to use stainless steel for EO duty and to simplify the process as far as possible. 4.1.2. Ethylene oxide plant explosion at BP Chemical Antwerp on 3 July 1987 (Melin, 1991) In this accident, an EO plant was destroyed by a violent explosion originating on the EO purification column. The cause of the explosion is a leak of EO from a manhole flange into the insulation packing at the base of the column. PEG formation and oxidation with air, produced a hot spot which initiated EO decomposition in a stagnant area inside of the column. The lessons learned from this accident are similar to those of the Basant accident, In addition, it was considered not insulating large flanges for EO duty. 4.1.3. Ethylene oxide plant explosion at Seadrift Texas, on 12 March 1991 (Viera, Simpson & Ream, 1993) The explosion of an oxide redistillation still was carefully investigated. It was shown that after a tube dryout, EO vapour could react with rust deposits at the temperature of the steam used, to produce a hot spot to above 500°C which could initiate EO vapour decomposition. 4.2. Influence of the vessel material of construction and of impurities on the thermal stability of EO The thermal stability of EO in the presence of various construction materials was investigated by different authors to check the compatibility of these materials with EO. Also of interest are studies on the thermal stability of EO in the presence of impurities like water or caustic soda. Of particular interest are thermal stability studies carried out using the Accelerating Rate Calorimeter (ARC). Freeder & Snee (1988) studied the thermal stability of 1.8 g samples EO mixed with 0.2 g aqueous sodium hydroxide solutions with mole concentrations between 0.125 and 1.0 NaOH. The polymerisation exotherm onset detection temperature was found to rise from 2.5°C for 1.0 NaOH mole concentration to 55°C for 0.125 NaOH mole concentration.

45

The thermal stability of pure and contaminated EO was investigated by Britton (1990) also in ARC tests. Pure EO was found to exhibit the threshold self-heat rate of 0.02°C/min at 210–220°C in titanium bombs and at 200°C in Hastelloy bombs. In the presence of water, the ARC threshold self-heat rate of 0.02°C/min was obtained at 150°C with 10% wt. water, 70°C with 40% wt. water and 50°C with 90% wt. water. See also the curve in Britton (1991). EO with 1% wt. dimethylamine exhibited a threshold onset temperature of 33°C. Ammonia is known to react with EO as in the reactors to manufacture ethanolamines. Therefore, the presence of water, amines and ammonia, and any other alkaline contaminant should be excluded in bulk EO storages. Rust is known to induce slow EO polymerisation at storage temperature and is supposed to have initiated a runaway polymerisation at higher temperature (Kletz, 1988). High surface area iron oxide caused the initiation of EO decomposition with an onset temperature of 98°C. So did aluminium chloride. Chlorides of Sn, Ti and Fe are said to reduce EO thermal stability. ARC tests on a charge of 2 g EO and 2 g insulation materials have been performed. A threshold heat-rate of 0.02°C/min was found at 70°C for asbestos, 85°C for calcium silicate insulation, 120°C for expanded perlite and 150°C for mineral wool. 4.3. Fires in insulation materials Fires in insulation materials was a frequent cause of explosion in EO production plants and ethoxylation plants. Leaks of EO in the insulation material have been shown to form polyethylene glycols (PEG). The insulation material and traces of rust are known to catalyse EO polymerisation at temperatures below 100°C (see above). Very divided PEG in insulation material can then oxidise and self-heat to above 500°C—a temperature high enough to initiate the violent decomposition of EO on the process side. The key literature references on this phenomenon are the papers of Britton (1990, 1991) and Schliephake et al. (1992) related to the Basant explosion. In the latter paper it was shown that a leak of EO in wet glass wool would give PEG at temperature down to 80–120°C, especially if the insulation material is impregnated with Fe203+1% CrO. Fifty percent of the PEG produced has a chain length of four monomers or more. These PEGs could oxidise in glass wool and selfheat to above 600°C, a temperature high enough to initiate EO decomposition. The practical conclusions on the experimental results found in the literature are that a construction material which does not rust would be preferable for EO duty, i.e. stainless steel. The best choice for the insulation material is foam glass with closed porosity, which would not enhance PEG oxidation. As pointed out earlier, large

46

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

flanges should not be covered by the insulation to avoid leaks into the insulation material. The choice of flanges and gaskets is very important. Top quality equipment should be used. Gasket materials which react with EO at process temperature should be excluded. Gaskets which would swell or yield at the process temperature should not be used. Gaskets which would immediately be destroyed in case of fire should be avoided. In European countries, spiral wound graphite gaskets are installed, which provide a good resistance in case of fire. If composite gaskets are used, one should check that components of the gasket do not dissolve in EO over time. 4.4. Choice of construction material for EO storage The above considerations may result in decisions made concerning the design and construction material for EO storages. Depending on the date of construction and on the location, the design of EO storages may be quite different. Probably, large storage vessels near EO production plants or even in ethoxylation facilities are made of carbon steel to lower their cost of construction. Carbon steel vessels should be cleaned and treated to exclude the presence of iron oxide which would promote the formation of PEG. For the same reasons and for safety considerations, these storage vessels are cooled and insulated. From time to time, it may be necessary to remove the insulation for corrosion and thickness inspection. If the construction material is stainless steel, both the refrigeration and the insulation are less necessary. There is less need for inspection and inspection is easier. The storage vessels are smaller and located in a pool, completely submerged with water or partly submerged with their top cooled by a water spray. Old storage vessels made of carbon steel kept under the sun and without refrigeration are of a lower standard. 4.5. Specific causes of runaway reaction 4.5.1. The runaway polymerisation/decomposition of an EO storage due to a back-flow of the ethoxylation reaction mixture from the reactor to the storage vessel This extremely serious accident must be prevented by redundant back-flow prevention, the most effective of which is to provide an EO storage pressure greater than the ethoxylation reactor operating pressure. This type of accident has occurred in plants manufacturing ethanolamines where the operating pressure may be well above the storage usual pressure, also ethanolamines and ammonia are very effective catalysts of the polymerisation of EO. In the Doe Run accident (Troyan & Levine, 1968) the ethoxylation reactor was a tube reactor where EO was fed in the liquid phase, thus making the back-flow possible. In the wide spread semi-batch ethoxylation pro-

cesses the provision of an EO feed in the gas phase and not through a dip pipe in the liquid would prevent a significant back-flow of liquid reaction mixture to the EO storage vessel. A temperature below the ambient in EO storage vessels could possibly prevent a runaway reaction in case of reaction mixture back-flow to the EO storage vessel since most contaminants would cause EO to react, only at temperatures above the ambient. 4.5.2. Contamination of EO by impurities The poor quality of EO delivery may lower its thermal stability and induce self-heating. The most probable contaminants are water, rust, glycols, PEGs. if the storage temperature is low enough, the impurities will not induce any detectable self-heating, the best solution is to consume the EO stock as soon as possible. Monitoring of the storage temperature to detect self-heating is only effective for significant temperature increases well above the daily variation of the storage temperature which may be of a few degrees Celsius. A good quality control would be more effective to detect contamination of the EO delivered.

5. Formation of nitrogen trichloride in chlorine manufacture Most chlorine is manufactured by NaCl or KCl brine electrolysis. Due to the presence of ammonium ions and fossil amines impurities in the salt and to the use of nitrogen containing flocculants for salt processing, nitrogen trichloride is present in crude chlorine. Nitrogen trichloride is formed by chlorination of ammonium ions and nitrogen containing organic compounds. Nitrogen trichloride is an unstable and detonating compound discovered by Pierre-Louis Dulong (17851838) with a theoretical atmospheric boiling point of 71°C. For a review of the dangerous properties of NCl3, see the thesis of Baillou (1990). In the chlorine purification process, nitrogen trichloride accumulates in a washing column where gaseous chlorine is washed with liquid chlorine under atmospheric pressure, i.e. at a temperature of ⫺35°C. Very frequently carbon tetrachloride is added to the column to extract NCl3 from liquid chlorine in the column bottom. The solution of NCl3 in CCl4 is removed from the column and nitrogen trichloride is destroyed either thermally by raising the temperature of the NCl3/CCl4 solution or by disposal of the solution. The control of the NCl3 concentration in the CCl4 solution is critical since NCl3 may decompose violently. A runaway reaction accident is described on this process in a pamphlet by the American Chlorine Institute (1975). An off-line reboiler used to decompose NCl3/CCl4 solution exploded on 17 October 1967 in a PPG facility at Lake Charles, Louisiana. The estimated NCl3 concen-

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

tration was 8.5% wt. in CCl4. Another incident on the same type of installation at an unspecified location was also reported in 1995. The DTA thermogram of the NCl3/CCl4 solution given in the American Chlorine Institute (1975) is typical of an autocatalytic decomposition. Similar DTA thermograms may be obtained in titanium or stainless steel closed cells but they differ from DTA thermograms obtained in sealed glass ampoules (see Fig. 6). The decomposition of NCl3/CCl4 solutions should be active in a temperature range of 40–60°C but may present an isothermal induction period depending on the wall material. The reaction is expected to assume a radical chain mechanism. The heat of decomposition of NCl3 in CCl4 solution is ⫺54.7 kcal/mol and 1 mol of nitrogen and 3 mol of chlorine are produced per 2 mol of NCl3 decomposed. Therefore, a continuous destruction process should be preferred to a batch destruction process based on thermal decomposition. Alternative processes to eliminate NCl3 are as follows. 앫 The withdrawal of liquid chlorine from the washing column bottom containing NCl3 for specific usage. This practice may also cause problems. 앫 A photochemical process to decompose NCl3 which may also present some drawbacks. 앫 A pretreatment of the brine to eliminate nitrogen containing impurities. Trace concentration of NCl3 is also present in bulk liquid chlorine shipped to customers. Following Eurochlor recommendations, the NCl3 concentration should not exceed 20 ppm in 1000 kg containers, 10 ppm in 20 000

Fig. 6. DTA thermogram of a 23.7 mg sample of 13.2% wt. NCl3 carbon tetrachloride solution measured in closed titanium test cell under 4°C/min temperature scan conditions. The exotherm pattern is similar to the thermogram shown in Baillou (1990).

47

kg and 50 000 kg rail cars and 2 ppm above 300 000 kg inventory (Eurochlor, 1990). The withdrawal of Cl2 from the gaseous phase is forbidden for containers of 1000 kg and more to avoid concentrating NCl3 in the vessel. The accumulation of NCl3 in chlorine vaporisers operating continuously is possible if the operating temperature is too low. If the vaporiser operating temperature allows enough NCl3 decomposition, NCl3 accumulation does not occur. The recommended frequency of NCl3 control in Cl2 production is given in Eurochlor (1990) depending on the observed concentration. For further information on the hazard induced by NCl3, see Dokter (1985), Gustin & Fines (1997) and Gustin (1997b).

6. The hazards induced by the presence of hydrogen sulphide and mercaptans in petrochemical feed-stock Petroleum gas as well as crude oil contain hydrogen sulphide and organic sulphur compound impurities. The combustion of non desulphurised oil will produce SO2 emission to the stack, making the use of low sulphur grade fuel more environment friendly. Petroleum gas processed in the petrochemical industry must be submitted to desulphuration for more specific reasons, i.e. the sulphur containing impurities would damage process catalysts, especially precious metal catalysts. Desulphuration may be carried out by passing the process gas over molecular sieves to remove water and sulphur containing impurities. Regeneration of the molecular sieves produces offgases containing hydrogen sulphide, mercaptans and organic sulphides. By passing these off-gases through carbon steel pipes and equipment, kept all the time oxygen free, ferrous sulphides are formed which are pyrophoric, i.e. they produce smouldering deposits if air is allowed to enter the process. This is a very dangerous ignition source for gas phase explosions if the vessel atmosphere is flammable. The formation of pyrophoric ferrous sulphides was described by Davie, Nolan & Hoban (1993) and Davie, Mores, Nolan & Hoban (1993) in papers on possible ignition sources in the storage of heated bitumen. It was shown that pyrophoric iron sulphides are formed under anaerobic conditions by passing hydrogen sulphide over FeIII rust deposits. That ferrous sulphide can oxidise under ambient temperature is not well known since it is necessary to heat pyrite ore to very high temperatures to produce SO2. However, pyrite dust gives dust explosions in sulphide mines (Reid, 1958; Liu & Katsabanis, 1993), thus proving the oxidability of metal sulphides under ambient temperature. It is unfortunate that these dangerous

48

J.L. Gustin / Journal of Loss Prevention in the Process Industries 15 (2002) 37–48

properties of sulphur containing petroleum feed-stock are not well documented in the literature even in extensive monography on the chemistry of sulphur (Reid, 1958). 7. Conclusion The presence of trace impurities in chemicals is a hazard when there is a potential for fast and exothermic decomposition or polymerisation reactions. Typical examples of these process conditions are documented in this paper concerning organic nitrocompounds and two reactive monomers vinyl acetate and ethylene oxide. These examples are quite representative of other unstable chemicals or reactive monomers. The detrimental effect of impurities in heavy chemical feed-stock has been examplified by two well known examples, the presence of nitrogen trichloride in chlorine manufacture and the presence of hydrogen sulphide and mercaptans in petroleum gas and crude oil. These two examples are also representative of common problems appearing in the chemical industry. References American Chlorine Institute (1975). Nitrogen trichloride. A collection of reports and papers. Report no. 21. (2nd ed.). New York: American Chlorine Institute. Anon. Encyclopedia of Polymer Science and Engineering (1988). 2nd ed., 12, 509. Anon. Maximum levels of nitrogen trichloride in liquid chlorine, Document GEST 76/55, 9th ed., September 1990. Brussels: Eurochlor. Anon. (1980). Les me´ langes explosifs. INRS. Anon. (1989). Explosion in a dinitrotoluene pipeline. I. Chem. E. Loss Prev. Bull., 88, 13–16. Anon. (1991). Explosion at the BASF Antwerp Ethylene Oxide/Glycol plant. Loss Prev. Bull., 100, 1–12. Anon. (1994). The fire at Hickson and Welch Ltd, 21 Sept. 1992. HSE Books. Baillou, F. (1990). Proprie´ te´ s explosives du trichlorure d’azote gazeux, thesis dissertation, 27 septembre 1990, Universite´ d’Orle´ ans France. Bateman, T. L., Small, F. H., & Snyder, G. E. (1974). Dinitrotoluene pipeline explosion. A.I.Ch.E. CEP Loss Prev., 8, 117–122. Britton, L. G. (1990). Thermal stability and deflagration of ethylene oxide. Plant/Operations Progress, 9 (2), 75–86. Britton, L. G. (1991). Spontaneous fires in insulation. Plant/Operations Progress, 10 (1), 27–44. David, D. J. (1965). Analytical chemistry, (vol. 37). pp. 82–85. Davie, F. M., Mores, S., Nolan, P. F., & Hoban, T. W. S. (1993). Evidence of the oxidation of deposits in heated bitumen storage tanks. Journal of Loss Prevention in the Process Industries, 6 (3), 145–150. Davie, F. M., Nolan, P. F., & Hoban, T. W. S. (1993). Study of iron sulfide as possible ignition source in the storage of heated bitumen. Journal of Loss Prevention in the Process Industries, 6 (3), 139– 143. Dokter, T. (1985). Fire and explosion hazards of chlorine-containing systems. Journal of Hazardous Materials, 10, 73–87.

Freeder, B. G., & Snee, T. J. (1988). Journal of Loss Prevention in the Process Industries, 1, 164–168. Gickel, D. (1983). La re´ action dangereuse de de´ composition des de´ rive´ s nitre´ s en pre´ sence de bases. Me´ moire CNAM. Centre du Haut Rhin, Universite´ de Mulhouse, France. Grewer, Th. (1994). Thermal hazards of chemical reactions. Industrial Safety Series, Vol. 4. Elsevier. Grewer, Th., & Klais, O. (1988). Exotherme Zersetzung, Schriftenreihe ‘Humanisierung des Arbeitslebens’, Band 84 VDI Verlag. Grewer, TH., & Rogers, R. L. (1993). Exothermic secondary reactions. Thermochim. Acta, 225, 289–301. Gustin, J. L. (1995). Assessment of runaway reaction hazards using the experimental methods of the process safety laboratories. In International Symposium on Runaway Reactions and Pressure Relief Design, Boston, 2–4 August 1995 (pp. 150–185). Gustin, J. L. (1997a). Runaway reaction hazards in processing organic nitrocompounds. I.Chem.E. Symp. Ser., 141, 393–403. Gustin, J. L. (1997b). Safety of chlorination reactions. In GEST 97/242, 4th Eurochlor Technical Seminar, 27–28 February 1997, Leipzig. Brussels: Eurochlor. Gustin, J. L. (1998). Runaway reaction hazards in processing organic nitrocompounds. Organic Process Research Development, 2 (1), 27–33. Gustin, J. L. (2000). Safety of ethoxylation reactions. I.Chem.E. Symp. Ser., 147, 251–263. Gustin, J. L., & Fines, A. (1997). Safety of chlorination reactions. I.Chem.E. Symp. Ser., 141, 405–416. Gustin, J. L., & Laganier, F. (1998a). Understanding vinyl acetate polymerization accidents. In First Internet Conference on Process Safety, 27–29 January 1998. Gustin, J. L., & Laganier, F. (1998b). Understanding vinyl acetate polymerization accidents. I.Chem.E. Symp. Ser., 144, 387–403. Kletz, T. A. (1988). Fires and explosion of hydrocarbon oxidation plants. Plant Operations Progress, 7 (4), 226–230. Kurland, J. J., & Bryant, D. R. (1987). Shipboard polymerization of acrylic acid. Plant/Operations Progress, 6 (4), 203–207. Levy, L. B. (1987). Inhibitor–oxygen interactions in acrylic acid stabilization. Plant/Operations Progress, 6 (4), 188–189. Levy, L. B. (1993). The effect of oxygen on vinyl acetate and acrylic monomer stabilization. Process Safety Progress, 12 (1), 47–48. Levy, L. B., & Hinojosa, L. (1992). Effect of oxygen on vinyl acetate polymerization. Journal of Applied Polymer Science, 45, 1537– 1544. Liu, Q., & Katsabanis, P. D. (1993). Hazard evaluation of sulphide dust explosions. Journal of Hazardous Materials, 33, 35–49. Macnab, J. I. (1981). I.Chem.E. Symp. Ser., 68. Melin, B. E. (1991). Ethylene oxide plant explosion, 3 July 1987. B.P. Chemicals Antwerp, Belgium. Loss Prevention Bulletin, 100, 13– 20. Nicholson, A. (1991). The effect of O2 concentration on methacrylic acid stability. Plant/Operations Progress, 10 (3), 171–183. Reid, E. E. (1958). Organic chemistry of bivalent sulfur, Vol. 1. New York: Chemical Publishing Co. Inc. Riethmann, J. et al. (1976). Chem. Eng. Tech., 48, 729. ¨ . (1992). Formation and Schliephake, V., Giesbrecht, H., & Loffler, U self-ignition of polyglycols in insulation materials: a source of danger for ethylene oxide plants. In Seventh International Symposium on Loss Prevention and Safety Promotion in the process industries, Taormina, Italy, 4–8 May 1992 (pp. 1–22). Paper no. 8. Stoessel, F. (1989). Sixth International Symposium Loss Prevention, Oslo, 1989, Paper 77. Tong, W. R., Seagrave, R. L., & Wiederhorn, R. (1977). A.I.Ch.E. Loss Prevention, 11, 71–75. Troyan, J. E., & Levine, R. Y. (1968). Loss Prevention, 2, 125–130. Viera, G. A., Simpson, L. L., & Ream, B. C. (1993). Lessons learned from ethylene oxide explosion at Seadrift, Texas. Chemical Engineering Progress, August, 66–75.