On the Catastrophic Explosion of the AZF plant in Toulouse
Septembre 21, 2001
Department of Chemistry, University of Tennessee,
Knoxville, TN, 37996-1600, USA
Prepared for Presentation at
8th Global Congress on Process Safety
April 1-4, 2012
AIChE shall not be responsible for statements or opinions contained
in papers or printed in its publications
On the Catastrophic Explosion of the AZF plant in Toulouse
On Septembre 21, 2001
Department of Chemistry, University of Tennessee,
Knoxville, TN, 37996-1600, USA
Keywords: Ammonium nitrate; accidental explosion; pollution and high voltage line.
During the morning of September 21, 2001, around 10:17 am, part of a warehouse containing about 300 metric tons of rejects of ammonium nitrate granulates exploded in the AZF/Grande Paroisse plant in Toulouse (France), a subsidiary of TOTAL, causing 31 deaths, hundreds of casualties, and enormous damages, estimated at over 2 billion Euros, to the plant, its neighbors and the whole city. The criminal investigation focused exclusively on the explosion itself and neglected the long chain of precursor events that took place in the previous hours, reported by numerous witnesses, and suggest that a complex cascade of successive events was triggered by some unknown event. These strange, unusual phenomena were never investigated. The prosecution adopted an unlikely chemical scenario that cannot explain the origin of the catastrophe. The official investigation and the first instance trial were inconclusive. The judge found insufficient ground for criminal liability. An appeal trial is in progress. A detailed examination of the different precursor events suggests a tentative explanation for the succession of events that linked to lead to the catastrophe. A strong thermal inversion caused pollutants from local plants to affect a high voltage electric line downwind from the plants, generating strong flames and arcs without triggering the current breakers and perturbing the computers that control the plants. Rogue orders caused a chain of explosions in the neighbor SNPE plant and finally that of a WW II bomb under AZF.
1.1 Origin and location of the “Chemical Pole”
A powder plant was built in the 19th century on the West side of the river Garonne, South of Toulouse, then far from the city, in a nascent suburb, about three miles from the center. It was transformed into a plant manufacturing nitrocellulose for artillery shells in 1914. After WW I, it was given the rights to the Haber process that France received from Germany and it began manufacturing ammonia and ammonium nitrate used first as a fertilizer. It expanded and, over time, it was sold from one company to another until it became a small, remote part of the TOTAL group as the AZF plant of the company Grande Paroisse. Toulouse, which might have been a 100,000 people city at the end of WW I, grew considerably and is now a half-million people city in an over one million people urban agglomeration. The plant, which was built in a remote suburb, became engulfed in the growing suburb. Attempts by the plant owners to move it away from the city was strongly opposed by the locals and failed. Now the area has been turned
Figure 2 Figure 1
In the 20th century, an industrial zone developed around the AZF plant. A large plant was in the East of it, on an island separated from AZF by a narrow, shallow, rather stagnant arm of the river. This plant belonged to the Societe Nationale de Poudres et Explosifs (SNPE), a state-owned company, part of the French Department of Defense, now sold to a private company. It manufactures rocket fuels and a few important reagents, including phosgene, sold to the nearby Tolochimie plant, South of AZF, which uses it as an intermediate in complex organic chemistry syntheses. A municipal cogeneration plant, burning garbage and selling electricity to the French grid managed by Electricite de France (EDF) is 1.5 mile away. A few minor plants, a large bus center, a major transformer and control center of the EDF grid and a small commercial center grew in the North-West corner of the plant. This group, locally known as the “Chemical Pole”, is separated from the city by a sort of high levy over which were built a local road, a highway (the “rocade”), a railway track, and the high voltage line (63 kV) that feeds the local plants. Figure 1 is a map showing the Chemical Pole. Figure 2 shows the two plants, south of the rocade.
1.2 The AZF plant
The AZF plant used to make ammonia after the Haber process and to oxidize part of it into nitric acid. The required hydrogen is made by controlled oxidation of methane into carbon monoxide and carefully purified. Ammonium nitrate (AN) is made by neutralization of nitric acid with ammonia, using the heat of neutralization to vaporize part of the water. Eventually, a solution of 98% AN in 2% water is heated at 200oC and sprayed at the top of a 50 m high prilling tower. The droplets fall in a counter-current of air aspired from the top of the tower. The solution loses most of its water. The AN particles collected in the bottom undergo further treatments to enlarge their average size and to cover them of small amounts of a limestone powder to reduce the trend of the particles to cake. At the top of the tower a complex structure prevents most particles from escaping into the cloud of condensed steam that forms when the air current mixes with cold air (Figure 3). Regulations stated that the prilling tower should not have lost more than 35 kg/hour of AN. The particles that escaped into the atmosphere capture water and turn into a brine of AN that, eventually, fertilizes the environment for no useful purpose and covers the roofs of buildings in the North of the plant with a white layer that is washed away by the rain.
Most of the serious problems encountered in the AN industry arise from its hygroscopicity. In humid atmosphere, AN captures water from the air; the particles get covered by a thin layer of saturated water solution. In dry air, water vaporizes and leaves a multitude of fine crystals between close particles, progressively welding them. Because AN is used as a fertilizer by spreading large amounts of the particles over wide fields, this caking effect must be controlled, otherwise much time would be lost in breaking down the cakes formed. Because production of AN is continuous while its sale is seasonal, large storages of the fertilizer exist everywhere. Large heaps of solidified AN used to be broken by explosives, which lead to a few major accidents or catastrophes (e.g., Oppau, 1921). The use of heavy fuel to coat particles proved effective but caused the catastrophes of Texas City (TX) and Brest (France) in 1947. Now the particles are coated with a thin layer of a limestone powder.
During production campaigns, some of the ammonium nitrate made has to be rejected. Mostly, this includes the particles that are too small or too large and are selected by appropriate sieves; the material produced during the beginning or the end of a campaign; the material that falls from the conveyors during transport across the plant; and the material produced when attempting changes in the composition of the fine layer of limestone. This rejected product was stored in a warehouse called “hangar 221” for periodic shipping to other plants making complex fertilizers, e.g., superphosphates. In spite of arduous efforts, the accusation could never prove that there was anything wrong with the reject products nor with the conditions under which it was stored.
2. The Toulouse catastrophe
Around 10:17 am on September 21, 2001, an enormous explosion took place in the hangar 221 of AZF where ammonium rejects were stored before processing for the preparation of mixed fertilizers. Thirty one people perished. Numerous others were wounded. The AZF plant was ravaged, numerous close by plants, companies, buildings were destroyed. Major material damages took place across the city, with a reported 50,000 windows broken, with the flying glass shards wounding many people, and a reported 2 billion Euros in material damages.
Figure 3 (above) shows aerial views of the plant taken in July (top) and early (bottom) October 2001. The lower photo shows the crater from the exploded hangar 221 and the conveyor carrying the product after finishing to the main storage area. The beams carrying the conveyor were painted green. They are not seen on the top figure because they are hidden by the top of the conveyor. The white spots on the roof of the main storage of the AN are deposits of AN coming from the prilling tower. Figure 4 shows the prilling tower before (left) and shortly after (right) the explosion of hangar 221.
Figure 4. The prilling tower before and
Figure 4. The prilling tower before and after
3. The official explanation
The official experts of the accusation and the instruction were exclusively concerned with the explosion of hangar 221. To explain it, they focused on a chemical reaction that they imagined could have happened although this seems highly unlikely if not totally impossible.
The AZF plant manufactures also DCCNa (dichlorocyanurate de sodium), a compound that is widely used to keep clean the water in swimming pools. Its hydrolysis produces hypochloric acid which can react with ammonium ions to give chloramines and, eventually, nitrogen trichloride.
Nitrogen trichloride is unstable and, if accumulated, it rapidly decomposes exothermally and may explode spontaneously when its concentration is sufficiently high. However, the plant was aware of the potential problem. The productions, transports, and storages of DCCNa and AN were kept well separated. No DCCNa was ever found on the grounds of the areas where storages of AN rejects were temporarily kept. No reaction can take place between the dry solids which could not be mixed if one product was poured over a heap of the other one. The hangar 221 was dry that day, the humidity of the dry Autan wind being low. This scenario was never even tried at any scale, except by finding (after 74 attempts that were not properly reported) the very exact conditions of humidity and mixing leading to a partial explosion.
4. The unusual events reported by numerous witnesses
During the morning of September 21, 2001,numerous witnesses reported many exceptionally unusual events. Significantly, although most of the details of these strange observations differ markedly from one report to the next, it is clear that the events reported were all quite similar in nature. Each fact reported was almost unexpected, very brief, very intense and localized, and it affected only a limited volume of space. Depending on the witnesses’ location, the direction of their lines of sight and their frames of mind, witnesses who saw the same phenomenon might have provided reports differing in many details. All of these reports were actually ignored by the investigators, the prosecutors, the lawyers, and the judges based on their apparent inconsistency, on the lack of reliability and sophistication of the witnesses, and on the impossibility for jurists to find a simple, straightforward explanation of these events. Yet, it is clear that these reports indicate the profound troubles that were taking place in the area that day and were getting progressively worse. Based upon these reports, we will conclude by suggesting a tentative explanation of the complex series of events that happened but should never have taken place.
4.1 The thermal inversion
The dominant wind in the region of Toulouse is a South-South-West to North-North-East wind named “Autan”. It was moderate that day (1). A strong thermal inversion was noted by the French weather service for September 20 and until late morning the 21st. Consequently, the atmosphere was stagnant all around the plants. When the river Garonne flows out of the Pyrenes mountains and enters the plain around Toulouse, between the AZF and SNPE plants, it is about 150 m below the Pech David plateau on its East side and about 50 m below its less steep left bank. This provides a sort of cauldron in which a polluted atmosphere may accumulate during thermal inversions, in which it may simmer and where chemical reactions may take place.
4.2 The extent of the local pollution
Numerous witnesses reported strong, unpleasant chemical odors around the plants, particularly on the North part of the industrial area. Figure 5 (next page) shows the distribution of these witnesses. Smells of ammonia and urine were stronger and more aggressive than usual or than ever before. During the morning, the industrial site became increasingly polluted. A witness saw a dense, hazy polluted atmosphere with a fog bank over the SNPE side of the river and above the chimney of this plant (2). Another one saw a thick, white plume of condensed vapor flowing from the prilling tower and drifting horizontally westward (3). This was likely a plume of condensed AN brine kept close to the ground by the thermal inversion and pushed by the Autan wind toward the high voltage line. Around 10:10 am, a witness (4) saw a light grey cloud,
Figure 5. Distribution of witnesses
reporting unpleasant chemical smells in the morning of September 21, 2001
Figure 5. Distribution of witnesses reporting unpleasant chemical smells in the morning of September 21, 2001
about 100 yards long, drifting along the stagnant arm of the river, below the highway bridge but above the railway track. A later report tells the same cloud to be longer, thicker, and higher a few minutes later (5), covering part of the highway, more than 100 to 200 yards long, and drifting toward the high voltage line.
4.3 The electric perturbations
The Marchand hospital, on the other side of the road from AZF, is a priority hospital that should never lose electric power. Yet, it completely lost power for 15 min. around 9:30 am (6). The generator of the SETMI cogeneration plant about 1.5 miles West-South-West of AZF was suddenly disconnected from the network, for a still unknown reason, between 10 and 20 seconds before the explosion of hangar 221 (7). The instruction and the trials refused to consider the detailed report provided by its director. Numerous local users reported abrupt losses of electric power (8-12). In the computer center of AZF (13), the fire extinguisher system was abruptly activated for no apparent reason about 10 seconds before the explosion of hangar 221.
4.4 The perturbations of local communications
Many AZF employees reported serious troubles with their phones and phone lines shortly before the explosions. These problems include abrupt interruptions of communications, strong electric shocks to users, sparks jumping out of office phones (13, 14). Two operators saw unexpected and unjustified alarms displayed on their monitors 30 sec before the explosion (15). Two experts stated that the data recorders went crazy long before the explosion of hangar 221 (16). They felt that the plant computers had gone out of control some time before the final explosion.
4.5 Unusual lights and light-balls
In Europe Sols, a shop facing AZF, two witnesses heard a first explosion, saw turbulent balls of fire rapidly passing by then heard the explosion of hangar 221 (17,18). Other witnesses saw flashes near 10:15 am and a straight, narrow light bolt exiting from the gate of the AZF plant near the high voltage line (19), or various lightnings (20-22), or a bolt followed by flames (23). M. Patrick Dupont, driving along the rocade, heard an explosion, stopped and saw a very brilliant lightning illuminating the plant behind a line of trees, shortly before a second explosion (24). Finally, a witness driving along the rocade saw a diesel locomotive on the rail track that was making the electric arcing noises of an electric one before hearing the explosion (25).
4.6 Partial and complete electrocutions
M. Maille was loading a tank car with ammonia (26). Suddenly, he heard a strong noise, like a thunderbolt close by. Startled, he places one hand on a nearby metal door, to keep equilibrium. He feels electrified, is immobilized, and shocked. A few seconds later, he heard a second explosion, is freed and falls. MM. Romero (27) and Dupont (28) are filling huge plastic bags of AN in a warehouse when their hands became suddenly stuck to the metallic funnel of their machine, they felt immobilized and electrified for a short while. The second explosion freed them. After the catastrophe, M. Jandoubi (29) was found dead in a large heavily damaged metal container that he was filling with AN bags brought to him by a conveyor. His corpse had extensive burns on his hands and torso and had a broken leg with an open break, but no bleeding. This suggests that he had died before his leg was broken, when the nearby hangar 221 exploded.
5. The mysterious smoke columns arising from the SNPE grounds
In a short period of time, several huge columns of smoke or dust arose suddenly from the ground in several places in the SNPE plant. Sixteen witnesses reported these columns but there were fewer than sixteen different columns (30). Using the directions and other indices provided by these witnesses, this study by Arnaudies demonstrated that all these columns took place in the Western part of the SNPE plant center, close to the stagnant arm of the river (see Figure 10, below). Four of these columns, those seen by MM Durand (31), Rizzato (32), Roux-Levrat (33), and Correnson (34) are well identified. They are important for different reasons.
5.1 The Durand column
Mr. Durand (31) was walking across the SNPE where he was working. Suddenly, he saw in front of him a large brown cloud rapidly moving toward him, between two buildings. This cloud was thick and made of ocher particles, as he realized later when he could brush his forehead. The cloud engulfed him, he was thrown back by 2 to 3 m and fell face to the soil. In the same time, he heard an explosion that he thought was gigantic. The pressure wave that he withstood was estimated at 150 mbar, not strong enough to wound him nor even to break his eardrums. When he turned over and got up, he saw a very high brown smoke or dust column that was rising most quickly, just against the chimney of SNPE. He rose and ran away. Later, he drew a schematic for the police (Figure 6, next page).
Figure 6. The Durand column
Figure 6. The Durand column
5.2 The Rizzato column
M. Rizzato (32) was working in the 4th floor of a building, about 20 m above ground, North of AZF and the rocade. Waiting for a load of building material to be brought to him by a crane, he was looking toward the AZF plant. Suddenly, he felt his building shake, heard a strong noise, and saw a wide, rather flat white cloud appearing on the SNPE ground and spreading. A short time later, a huge white, vertical column surged very rapidly in the middle of this cloud, straight as an I, and raised to a height that was later estimated to be several hundred meters (1,000 ft). A mushroom formed on the top of the column. The whole system was initially white but progressively turned ocher. Note that few combustible compounds burn while giving a white smoke or cloud, not a black one. Most probable would be hydrogen, ammonia, hydrazine and dimethylhydrazine, M. Rizzato drew schematics of the column and of its evolution (Figure 7).
5.3 The Roux-Levrat column
M. Roux Levrat (33) was driving toward Toulouse, between the Marchand Hospital and AZF, when he heard a strong explosion. He came to a stop and turned toward the plants where he saw a huge dark-grey column, roughly 10 m in diameter and 300 to 400 m high, which rose rapidly. Questioned later, he was sure that the column had formed in the SNPE ground but was unable to provide the location of this column inside this plant. After about eight seconds, the column collapsed while he heard a second explosion, that of hangar 221. He gave a schematic (Figure 8).
Figure 7. The stages of the Rizzato
Figure 7. The stages of the Rizzato column
5.4 The Correnson column
What M. Correnson (34) reported took place in a very short time. He was working in SNPE for a contractor, when he saw a bright white thunderbolt, heard a strong explosion and was violently thrown onto the ground by a strong burst of air. While getting up, he heard a second, enormous explosion and, immediately, saw a black column in the shape of a large mass arising very rapidly from somewhere in the direction of the explosion and then collapsing nearly as fast as it arose.
The direction that he provided was that of the hangar 221 but he could not positively identify this building as the origin of the black cloud because he could not see it from where he was, due to the presence of a building facing him in SNPE. This black column began as a smoke cloud, black in the center, orange on its edges and becoming dark red in its centre, with small black dots in the red region. The mushroom cloud of the explosion of hangar 221 quickly filled up the sky and Mr. Correnson fled away. He gave a schematic of what he had seen (Figure 9).
Figures 8 (left), the Roux-Levrat
column and Figure 9 (right), the Correnson column
Figures 8 (left), the Roux-Levrat column and Figure 9 (right), the Correnson column
6. Other strange observations
Several unusual, strange effects were observed that are inconsistent with the assumption made by the prosecutors that the catastrophe was the result of a single, simple explosion. The following list of unusual phenomena concerns the locations of the people who suffered burns besides their possible wounds, the distribution of some of the places having suffered unusually severe material damages, the various events that took place in SNPE besides the columns reported above, and the mysterious earthquake that was recorded at the time of the explosion.
6.1 The location of the people having suffered serious burns
One expert reported that most, if not all these people, were in the neighborhood of the high voltage line, on the North side of the AZF plant when they were burned (35). Figure 11 shows the distribution of these people (36-42). They were too far from the hangar 221 for these burns to have been caused by the heat wave generated by the explosion of the AN storage. The burns must have been somehow connected with their proximity to the high voltage line.
6.2 Important material damages to places remote from the explosion
A map of the distribution of the excess pressures caused by the shock wave due to the explosion of hangar 221 was drawn by INERIS, a local branch of the French environment administration (see Figure 12). These pressures were estimated from the material damages observed around hangar 221. Figure 12 shows the high estimates provided. The 20 mbar isobar curve passes slightly South of the André Daste school and far from Toulouse city hall. In both places as well as in a few other ones, the material damages seem to have been more significant than what is expected from this map, suggesting that they could have been due to some other causes. Hodin has used a conventional diagram to relate the material damages due to an explosion, the distance from this explosion, and the equivalent size of the TNT bomb that would have caused these damages (43). From a qualitative description of the damages experienced in one place, a reduced value, d, of the distance to the explosion is estimated. The ratio of the actual distance, D, to this reduced distance provides an estimate of the TNT-equivalent mass of exploded material (43).
Application of this method to a few typical cases provides surprising results. For example, in the school André Daste, close to the East bank of the main stream of the Garonne, Ms. Benyacoub, a schoolteacher overlooking a class pause, heard the noise of an electrical arc, saw a flashing light and was abruptly and brutally projected against the building wall by a very powerful burst of air. A large fraction of the school windows were broken and glass fell along the building (46). This suggests that the damages resulted from the explosion of 2460 tons of TNT while the actual explosion was estimated to be equivalent to that of approximately 40 to 50 tons of TNT. Slightly farther North, along the West bank of the Garonne, Ms. Roegel (47) attending a meeting 35 Bld des Recollets felt the building shake and heard a strong explosion that broke all the windows in the building. This suggests an equivalent TNT mass of 394 tons, again far too high. In City Hall, the mayor is on the phone and the line is interrupted. The mayor gave a dramatic description. The windows are torn from the walls and thrown into the middle of the rooms. This would correspond to the explosion of between 240 and 3000 tons of TNT, excessive in both cases.
Admittedly, the propagation of explosion shock waves in an urban landscape with steep hills, valleys, and many high buildings is erratic and interferences may cause damages more severe in some places than in others that are close by. However, these observations were repeated many times. They are troubling and make one wonder whether there were other explosions, like those of clouds of combustible gases mixed with air and emitted by one of the plants.
6.3 The damages experienced by the high voltage line
The high voltage line (63 kV) built along the highway north of the AZF and SNPE plants was cut at the time of the explosion of hangar 221. Curiously enough, there is no detailed record of the damages experienced by the pylons, their insulators, the cables and their supports nor by the transformers, controllers and computers associated with it. The available information is terse (46) suggesting that the magistrate in charge paid little attention to this question. Yet it is known that other damages took place besides the line being cut around the time of the explosion. Some local transformers were destroyed in the two plants and also in the immediate neighborhood. After the explosion, EDF replaced all the cables of the electric line, from the river to the west part of the AZF plant, south of hangar 221. Most of the recovered cables were sold to metal recycling outfits, without being treated as evidence. A few short sections were examined years later, at the request of lawyers. Their examination by experts showed most important clues ignored by the instruction, the accusation and the defense.
Short line sections that were near the cut had been kept. Their examination showed that the line was not cut by the fall of a heavy fragment thrown high with the explosion debris. The line had melted in many places. Some of the individual wires twisted into the cables had been broken and were then soldered together. Other parts of the cable were covered with many droplets of metal later solidified (46). Many of these filaments carry the traces left by electric arcs starting from them. The line was not cut but worn out by thermal erosion. This confirms that the many strange luminous phenomena recorded (see earlier) were related to this high voltage line. There is no information regarding the status of the pylons and insulators, which prevents an extensive study of how chemical pollution due to different possible sources might have caused these effects.
6.4 What happened at SNPE besides the columns?
The former director of the SNPE plant, acting in 2001, testified under oath that, as far as he knew then, nothing serious had happened in his plant before the explosion of hangar 221 in AZF and that it was only this explosion that ravaged his plant. This clear statement is in total contradiction with the reports of many employees or contractors of SNPE but it was not challenged in court. M. Jurado is walking across SNPE when he is suddenly thrown onto the ground (47). When he picks himself up, he sees the top of the prilling tower of AZF taking off as a rocket but the explosion of hangar 221 that immediately followed the destruction of the tower prevents him from seeing where the top of the prilling tower fell. He reports that the buildings around him are seriously damaged when he gets up, hence BEFORE the AZF explosion (47). Similarly, several other SNPE employees (48-51) clearly saw and heard two successive explosions. One witness begins by telling the policeman how his management had instructed him to say that he heard one explosion only, which he does. Having said that, he continues talking about what he was doing at the time of “the first explosion” and then “when the second explosion occurred”, a few seconds later. As reported earlier, MM Rizatto (32) and Durand (31) saw immense columns jumping from the ground close to them toward very high levels, nearly 1,000 ft high, before the explosion of hangar 221. Do not these witnesses confirm that it was not just a regular day at SNPE, until the AZF plant exploded?
Figure 13 below compares aerial photos of SNPE taken in July (left) and late September (right) 2001. Important material damages in SNPE are obvious. They cannot be explained by the explosion of hangar 221 that is located slightly North-West of the roofs seen on the photo.
6.5 The two explosions and their timing
Numerous but not all the witnesses reported having heard two successive explosions about 8-10 second apart, the first one shorter, sharper, and weaker than the second one. This raises two questions for which no consensus has yet been reached: Was there only one explosion generating two noises or were there two different explosions? If there were two, which one was due to the explosion of hangar 221 and what was the other one due to? The answers to these questions rapidly became obscured by serious political issues.
Within hours after the catastrophe, the French President, Jacques Chirac, and the Prime Minister, Lionel Jospin, flew to Toulouse to confer with the local Governor and Attorney General. The latter immediately declared to the press that he knew what had happened and would prove it shortly. He also stated that the rumor spread by hundreds of people who claimed to have heard two explosions was utterly wrong; there had been only one explosion, that of AZF. This had two opposite consequences: many people, particularly in the police, the justice and the state and local administrations took this statement as an order; many other people, including most scientists and many fans of American crime movies began smelling a cover-up. The police spread the story that explosion noises propagate simultaneously in the ground and in the air, explaining why witnesses could hear two noises from a single explosion. The problem is that ground vibrations are weakly coupled with air and are not heard. Furthermore, the distance between the two noises should increase in proportion to the distance between the witness and hangar 221 and remain constant at a given distance of the hangar.
Detailed investigations by French scientists provided an answer to the first question. Joets  asked to all witnesses having declared that they had heard two successive explosions what did they do in the mean time. He could find twelve witnesses, distant from hangar 221 by between 110 m and 48 km, who reported such actions and helped him to approximately time the duration between the explosions. The delay between them was found to spread between 6 and 11 seconds, to be independent of the distance from the hangar, but to depend on its direction. This clearly demonstrated that there were two distinct explosions and that these explosions did not take place in the same location. Arnaudies  and (53) analyzed records made during a few meetings taking place in Toulouse the morning of the explosion. He showed that two explosions had really taken place in different places and, from the time differences between the two explosions recorded in different place, he demonstrated that the first explosion did take place in the SNPE plant.
6.6 The timing problems
Actually, it turned out that it was far more difficult than expected to accurately estimate the time differences between events taking place in different locations with the desirable accuracy since the different plants involved used time lines that were not synchronized. No time line of events taking place in either company involved, AZF, SNPE and EDF, was synchronized properly with the universal time line of the Paris Astronomical Observatory. It proved impossible to tell which event took place first in the long series of events recorded in these companies. For technical reasons, also possibly for other ones, and because the judge did not seem to care, it is not known when EDF realized that there were unusual and possibly dangerous events taking place on the 63 kV line nor whether an automatic switch broke an electrical connection, whether power was restored, how often it might have been interrupted, nor what was the nature of the energy losses.
6.7 The earthquake and its mysterious record
A magnitude 3.4 earthquake is reported to have been recorded at the time of the explosion of 221 in several European locations. Strangely however, little detailed information about it was leaked out from these stations. What was the exact time when it took place? Where was the epicenter?
There is no history of so large earthquakes in the Toulouse area. It is unexpected to associate such a powerful earthquake to the surface explosion of 50 tons of equivalent-TNT. Recently, a 200-ton storage of declassified Swiss military ammunition that was buried deep in a tunnel in the Alps exploded accidentally. This explosion was recorded as a 3.7 magnitude earthquake. The explosion of a train in Ryongchon (China-North Korea border) on April 4, 2004, was explained by that of 800 tons of AN. It gave a 3.6 magnitude earthquake recorded by the Comprehensive Test Ban Treaty Organization. The magnitude of the earthquake due to the first Korean nuclear explosion was only 3.4. Finally, a 25 ton truck of AN had a traffic accident in Barracas, 40 miles North-West of Valencia, Spain, on March 9, 2003. Gas oil leaked from its tank and burned with production of an intense black smoke, with reddish shades due to the nitrogen oxide produced by dissociation of the nitric acid. A crowd gathered to watch this fascinating fire. After about 20 minutes, the load exploded, killing a dozen and generating an earthquake of magnitude 1.5. If a 3.4 magnitude earthquake took place in Toulouse and is not due to the explosion of hangar 221 in AZF, where could it have taken place?
The only record of the earthquake ever published is shown in Figure 15 . It shows three groups of peaks. The official interpretation, supported by the prosecution, is that the first group (P1, P2) corresponds to vertical motions, the second to horizontal motion (P3, P4) of waves conveyed by the ground surface and due to the explosion of hangar 221. The last signal (Ac) would be the acoustic wave of the sound of the explosion that moves more slowly. Consistent with the request of the local prosecutor, but without any scientific proof, the equipment not being linked to the international time base, the person having made this record attributed the first peaks to signals originating from the explosion of hangar 221, the little peak at the end of the record being due to the noise of the explosion. This is all the more surprising because the instrument was poorly connected to its sensors, poorly sensitive and not suited for recording air waves. An alternate, more likely explanation provided during the trial would consist in considering that the first group of peaks on the record, which is the one that certainly corresponds to the magnitude 3.4 earthquake, is due to the explosions that took place in SNPE and caused the columns reported by MM Durand, Rizzato, and Roux-Levrat while the little peak at the end, about 10 second later, is due to the ground vibrations caused by the explosion of hangar 221. This would mean that the explosion or explosions taking place in SNPE would have been underground explosions, as such known to be strongly coupled with the ground.
The conclusion of the prosecution that one single event might have caused the catastrophe of Toulouse is unacceptable. It is impossible that such a complex series of intricate phenomena could have one single cause. It is far more likely that this disaster is due to the evolution of a series of processes, one causing the next one in a sort of chaotic chain of processes, until the final ruin of the plants stops this chain. A proper explanation must incorporate, if not all at least most events reported and should provide, if not a detailed mechanism for each one of them, which would be impossible, but at least a coherent chain linking this series of unrelated processes, one triggering the next one while each one evolves in its own course, independently of the development of the following processes, without actual correlation between these independent processes. This complex chain of events might seem unlikely when first described.
Ammonium nitrate is generally a very safe product. Nearly ten million tons are produced yearly and few accidents have ever been reported. AN is a safe explosive. Its accidental explosions have only two known causes : (1) A sufficiently large amount of a powerful explosive is detonated in the immediate proximity of a large mass of AN. This is how the dramatic catastrophe of Oppau took place in 1921. A few other accidental explosions e.g., (Tessenderloo, Belgium) took place under similar conditions; (2) A large amount of ammonium nitrate is placed in contact with an organic material which is burning, as happened in Texas City (1947) and in Brest (1947) and more recently when large truckloads (about 25 tons of AN) were involved in traffic accidents, the gas-oil container of the truck was torn, gas oil leaked and burned, as in Barracas (Spain) and Buzau (Romania). This second cause was notoriously absent in Toulouse.
Some have suggested that terrorists could have deposited a sufficient amount of explosives in hangar 221 and triggered its explosion. The police certainly did not properly follow the trails of several suspects and neglected many possible leads. The main problem with a terrorist scenario, however, is that it cannot explain any of the precursor events reported here, nor the damages to the electric line, nor the explosions that took place in SNPE because there is no simple link between these phenomena that developed before the explosion of 221 and the acts of a terrorist.
Prosecution states that a small amount of sodium dichlorocyanurate (DCCNa) could have been poured on the heap of the daily AN rejects, which was stored in an anteroom of hangar 221, where it was brought by small tip-lorries, to be later incorporated into the major storage by a large bulldozer that scrapped the anteroom pavement. Unfortunately, as explained in one of the first sections, the scenario of the prosecution does not make scientific sense. It failed during the first instance trial. It is now crumbling in the Appeal Court.
The long series of serious incidents and accidents that took place in Toulouse during the morning before the catastrophe must be examined carefully, considering these events separately and successively, in their chronological order, in an attempt to understand what was their origin and how each one might have evolved to become the cause of the next one. We attempted to do that.
1- It is certain that during the night and the morning before the catastrophe, the neighborhood of the plants was the sites of a strong thermal inversion. The usual pollutants stagnated close to the ground. As reported by several witnesses, the odors from the plants were unusual in nature and stronger than ever before. So, the concentrations of these pollutants was unusually large (M. Merignac (2), with an important plume of droplets of a solution of AN coming from the prilling tower toward the high voltage line (3), and large banks of clouds of unknown composition drifting northward, along the branch of the river between the two plants. Statements made during the appeal trial have suggested that large amounts of uDMH (1,1-dimethylhydrazine) were accidentally or voluntary released from SNPE into the river between the two plants.
2- Through yet unidentified reactions, mixtures of air and this exceptionally high concentration of pollutants might have reacted together in the vicinity of and/or in immediate contact with the high voltage line while the cable, the pylons and the insulators of this line might have been coated with a layer of AN solution. Whether one of these pollutions played the essential role or a synergetic effect took place, the high voltage line certainly leaked energy into its environment, causing the flashes, sparks, arcs, and lights seen by many witnesses. The physical state of the surface of the short sections of this line that were recovered confirms this assumption (46). The line had certainly been heated and melted in parts, due to arcs and sparks that could only have been caused by unexpected chemical reactions on or close by the 63 kV line.
3- These phenomena explain also the electrification and the temporary or deadly electrocutions of employees at AZF (26-29). They also explain why many perturbations of the control systems of both plants must have taken place and why the emission of rogue orders is probable (if the controls of the AZF plant went out of order some time before the explosion (15,16).
4- In the same time and most probably triggered or otherwise related to these phenomena, several major accidents took place at SNPE. The connection between these three series of events, those taking place at AZF, those happening around the high voltage line of EDF, and the poorly documented events occurring at SNPE is uncertain. It might be that the arcs, flashes and flames around the high power line generated serious perturbations of the computers and control systems at SNPE as they certainly did at AZF and that some rogue orders sent by perturbed computers opened valves and released large amounts of flammable gases or aerosols in the underground corridors and caves that are known to exist in the island where SNPE was built (they were dug by the German army during WW II). It might be that a leak of uDMH had independently lead to the release of other flammable mixtures or that unexpected accidents took place when the SNPE co-generation unit was started, as it seems that it was planned to do some time that day. As a consequence, one or a few successive explosions took rapidly place in the SNPE underground, generating the columns of smoke and dust reported (31-33) and causing the earthquake.
5- The serious earthquake caused by this or these underground explosion triggered the explosion of one or two bombs dropped by RAF in 1944 on the plant. A photograph of the hangar 221, which already existed in 1944, was taken at that time. It shows a circular hole in the roof but the hangar was otherwise intact and survived another 57 years. It is known that about 15% of these old RAF bombs did not explode when they hit the ground. They buried under ground and, if not found and eliminated, remained there inert for many years. However, their explosion may be triggered any time by some significant perturbation like those caused by all kinds of earthworks. The accident that happened on June 1, 2010, in Göttingen (Germany) is a good illustration of the problems caused by these residues of the past. Workers found a bomb, called a mine-clearance squad who neutralized the first bomb and evacuated it. In the process, they found a second bomb, brought a robot to help them move it for easier access, but this second bomb detonated, killing three squad members. In the aftermath of WW II in France, earthmoving vehicles were rare and bombs buried deep in the ground were easily overlooked and forgotten.
Figure 16. The black trace
Figure 16. The black trace
This scenario is supported by a last observation, the black mark that was seen on aerial photographs taken early after the catastrophe (Figure 16) but was soon destroyed when the crater was modified for unknown purposes, upon the order of an unknown person. This last event is consistent with the general observation that AN is a very stable compound that explodes only when a significant amount of a powerful explosive is detonated in its immediate proximity .
8 - Acknowledgments
We thank all those who helped us during this investigation, particularly Laurent Jacob who provided many thoughtful suggestions, Daniel Montaron who provided numerous useful documents, much information that it was difficult for us to collect from afar and helped us in finding access to the trove of reports in the instruction file. A few still active, former or retired employees of AZF and SNPE, and other inhabitants of the Toulouse area have also been most helpful but they prefer to remain anonymous.
9 - References
 A. Joets, C. R. Geosciences, 341 (2009) 306-309
 J. M. Arnaudies, Natures, Sciences, Societes, 13 (2005) 421-425
 A. Souriau, M. Sylvander, V. Maupin, J. F. Fells, A. Rigo, C. R. Geosciences, 334 (2002) 155-161
 L. Medard, “Les Explosifs Occasionels”, Editions Technique et Documentation, Paris, 1987
The other references (shown as Dnnnn) are all from the instruction file of the trial in the Toulouse Tribunal of first instance. Most of them are to documents that were written by a police investigator under dictation by the witness, were reports of police interviews of a group of people, or, in some cases, documents received by the secretary of the judge in charge of the instruction. They are available from the Court and provided to the public as a giant file (ca. 6 GB) which is a collection of pdf files and can be downloaded. These documents are all written in the French language. Many are hand-written but most are accompanied with a typed version also.
(1) D2492, Meteo France report for 09/21/2001
(2) D596 p.2, Mr. Merignac
(3) D5443, Mr. Cabaret
(4) D593, Mr. Lavigne
(5) D601, Mr. Fourtet
(6) D2034, p.3, Mr. Palluel
(7) D4263, p.2, Mr. Haillecourt
(8) D4263, Mr. Hanoun
(9) D5010, p.85 Mr. Dumas
(10) D1269, Mr. Roques
(11) D4081, p.1, Mr. Pontillon
(12) D884, Mr. Adelando
(13) D3360, p. 3
(14) D368,1; D2110,p.1-2, Ms. Auzer.
(15) D4268, MM. Denis and Gamba
(16) D2972, p.8, MM. Donio and Znaty
(17) D4152, p.1, Ms. Gioan,
(18) D4154, p.1, Ms. Goudou
(19) D2483, p.24, Ms. Foiban
(20) D4409, p.2, Mr. Baby
(21) D3736, p. 61, MM. Ferrain and Dantras
(22) D3736, p. 61-2, Ms. Bordas
(23) D3736, p. 64, Mr. Sentenac
(24) D3736, p. 63 and D2043, p. 2, Mr. Patrick Dupont
(25) D606, Mr. Grau
(26) D5499, p.1, Mr. Maille
(27) D2109, p. 2, Mr. Romero
(28) D2111, p. 2, Mr. Dupont
(29) D82, Autopsy of Mr. H. Jandoubi
(30) D6399, p. 8, Mr. Arnaudies
(31) D5354, Mr. Durand
(32) D3955, Mr. Rizzato
(33) D5671, pp. 79-80, Mr. Roux Levrat
(34) D3353, p. 43-45 and D2803, Mr. Correnson
(35) D5671, p. 68, Mr. Hodin
(36) D5479, Albert
(37) D5481, Borgues
(38) D5481, Combes
(39) D5482, Compte
(40) D5483, Derrien
(41) D5484, Hebert
(42) D5414, Pascal
(43) D5671, p. 91, Hodin
(44) D5384, Benyacoub
(45) D5010 p.109, Roegel
(46) D3814, Francois
(47) D3353, p. 42, Jurado
(48) D3353, P. 46, Dessacs
(49) D890, Garrigues
(50) D3353, p. 45, Borderie.
(51) D3353, p. 47-49, Viguier
(52) D6595, Arnaudies