Holistic Case Study on the Explosion of Ammonium Nitrate in Tianjin Port

Abstract

On 12 August 2015, Tianjin port, Tianjin City, China, a catastrophic explosion of Ruihai International Logistics Co., Ltd. (Tianjin, China) killed 173 and hurt almost 798 people, accompanying a financial loss of almost USD 2 billion. The ignition of the first fire due to the autocatalytic decomposition of nitrocellulose was verified by differential scanning calorimeter (DSC) isothermal tests. A crater with a diameter of 97 m was created by the second explosion. For the second catastrophic explosion, an amount of 577 tons of trinitrotoluene was determined by the average through scaling law, crater inverse analysis and blast effects on structures. The overpressure against distance for consequence analysis was conducted using Baker’s, Sadovski’s and Alonso’s methodologies. A distinctive scenario of “two-successive-sympathetic detonations-following-a-fire” was proposed and discussed. Isothermal time-to-maximum-rate was validated to be approximately 9 days for the nitrocellulose inside the containers with an internal temperature of 60 °C stored at Tianjin port. A fatality radius chosen at the overpressure of 0.6 bar was ascertained to be nearly 410 m from the explosion origin.

1. Introduction

1.1. Occurrence of the Incident

At 22:51:46 on 12 August 2015, the warehouse of the Ruihai company for storing the hazardous materials, located at Jiyun road, Binhai new district of Tianjin port was first reported to catch fire in a container [1]. A container of dry nitrocellulose (NC) accompanied with spontaneous combustion was suspected to be the scene of the fire in the beginning. Thermal radiation and burning flame were the significant conditions that sped up the rapid fire spread throughout the storage area. In the summertime, because the container did not easily dissipate heat, the temperature of NC continued to rise to its auto-ignition temperature (AIT) and spontaneously burned. Infelicitously, the fire kindled by NC ultimately propagated to the containers of ammonium nitrate (AN) causing the consecutive explosions. Particularly, the second explosion caused the most catastrophic consequence. During the firefighting, the first explosion happened at 23:34:06 and the second larger explosion proceeded at 23:34:37; a videotape filmed the amazing explosions with extremely destructive powers and sparkling flashes. It exhibited the features of two rapidly successive explosions, which occurred extremely quickly with a time interval as short as 32 s. Six large fires and more than ten small fire sites were caused after the second explosion. At 16:40 on 14 August, the fire at the site was entirely extinguished. Figure 1 shows the pictures of explosion region before and after the occurrence of the tragedy. According to the statistics in the public media, 165 deaths with 8 missing individuals and 798 hospitalized injuries were reported. Besides, hundreds of buildings were demolished and an economic destruction roughly USD 2 billion.

1.2. Major Incidents Caused by Ammonium Nitrate

Since the 20th century, no less than 40 severe explosions were related to AN. Eleven representative catastrophes ignited by AN are summarized in

Table 1. Eleven major incidents caused by AN.

Incident Number	Location	Date	Quantity (Tons)	Fatalities/Injuries	Reference
1	Morgan, NJ, USA	4 October 1918	1000	64/100	[2]
2	Oppau, Germany	21 September 1921	450	561/1952	[3]
3	Tessenderloo, Belgium	29 April 1942	150	>100/NA	[2]
4	Texas City (TX, USA)	16 April 1947	2300	>500/3500	[3]
5	Brest, France	28 July 1947	3000	30/>1000	[2]
6	Shenzen, China	5 August 1993	65	18/873	[4]
7	Toulouse, France	21 September 2001	300	30/3000	[3]
8	Ryongchon, North Korea	22 April 2004	NA	161/1300	[3]
9	Monclova, Coahuila, Mexico	4 September2007	22–25	28/130	[3]
10	Tianjin Port, China	12 August 2015	800	173/798	[1]
11	Beirut Port, Lebaron	4 August 2020	2750	204/>7000	[5]

NA: Not available.

.

Several explosion incidents with less fatalities and injuries which occurred in Toulouse (France) [6] and West (TX, USA) [7,8] were reported. In 2020, a disastrous explosion in the warehouse of Beirut port which storing 2750 tons of ammonium nitrate exploded [9,10]. Society has suffered such a loss of properties and at the sacrifice of many human lives, and the industry should learn from the historical explosions caused by ammonium nitrate [11,12]; however, we have not [13].

1.3. Thermal Hazards and Autocatalytic Decomposition of Nitrocellulose

NC is a thermally unstable compound because of the presence of -NO₂ group, which has been classified as a flammable solid by the United Nations Committee of Experts on the Transports of Dangerous Goods (UN-TDG) and U.S. Department of Transportation (US-DOT). NC was numbered as UN 2556 or UN 3040 [14]. US-DOT classified it as the class 4.1 flammable solid [15]. Commercial NC was wetted with alcohols or water to suppress the sensitivity to explosive properties. Since 1955, the thermal decompositions of NC were studied by chemical analysis and spectroscopy [16]. In these two decades, some incidents caused by unstable features of the NC attracted researchers’ interests to validate the thermal stability of NC. Considerable exertions have been contributed to the the reactive instability of NC by calorimetry, thermogravimetry analysis (TGA), mass spectrometry, and surface technique [17,18]. In particular, by calorimetry and TGA, the exothermic onset temperature was determined non-isothermally to be of approximately 182 °C [19,20]. The enthalpy change of thermal decomposition was detected as high as 2591 Jg⁻¹ [21]. A research review on the kinetics of NC at 50–500 °C was reported by Brill and Gongwer [22]. An autocatalytic reaction of first order owing to the slower reaction to the O-NO₂ homeless in the 100–200 °C range was proposed and discussed by Kimura [23]. Thereafter, the autocatalytic kinetics of the thermal decomposition of NC were explored by several laboratories [24,25,26]. Most of these previous studies of kinetic triplets were put forward, however, without extending to the application of hazard analysis in the chemical industry. A study of thermal explosion of NC was declared by Luo et al. [27] using the Ozawa-Friedman method and bench-scale thermal explosion test as well as the Frank–Kamenetskii theory. An upper critical ambient temperature (UCAT) of the nitrocellulose decreased from 132.5 °C to 96.4 °C relative to the ignition dimension increased from 0.1 m to 1.6 m [27]. In the official report of Tianjin explosion, the first fire is reported to have been initiated by the ignition of NC caused by the autocatalytic decomposition in container at an internal temperature of approximately 60 °C [1]. Nevertheless, it is a regret that there is no direct evidence or experimental data to support this hypothesis up to now. One of the important objectives in this work is to predict the auto-ignition of NC from 60 °C with an experimental approach.

1.4. Fire and Explosion Hazards of Ammonium Nitrate

Ammonium nitrate has been used to make explosives and fertilizer since 1910 [28]. Neat AN cannot be ignited at room temperature; however, it can burn with the inductions of fire and contaminant. The oxygen balance of AN is approximately 20%, which reveals that the excess oxygen content serving as a donor for combustible materials. There are two commercial AN, ammonium nitrate of fertilizer grade (FGAN) and technical grade (TGAN) which hold the common chemical structure. In 2013, a fire happened at an AN plant in West (TX, USA) that eventually led to a serious explosion resulting in 15 deaths [13,29]. Nevertheless, in 2014 a similar event in Athens (TX, USA) only burned the warehouse out without further explosion [30]. A multiplicity of AN behaviors under different fire sites make it hard for the emergency staff to respond correctly. The National Fire Protection Association (NFPA) 400 classified AN as a Class 2 Oxidizer [31]. NFPA also assigned the AN as an instability rating of three to alert the personnel of emergency response that AN could be capable of detonating when confined at a fire site. The United Nations (UN) classified AN-Based Fertilizers (UN 2067) to be a Division 5.1 Oxidizer as well [32]. Besides, the United States Department of Transportation (US-DOT) followed this oxidizer classification. Being an oxidizer, it can facilitate the spread of fire and strengthen the combustion of contacted materials, even without air. Though AN is quite safe at ambient temperature, the thermal hazards of AN at high temperatures have been reported [33,34]. Contaminants, confinements, and external fire complicated the firefighting and enhanced the risk of AN stored in warehouse under fire [35,36]. An uncontrolled fire in particular will worsen the warehouse storing AN to cause an explosion with high probability [37].

1.5. Explosion Hazards and Susceptibility to Detonation of AN

Accidental explosions induced by AN were internationally responsible for several catastrophes. Not a few complete works had been executed by the U.S. Bureau of Mines in 1960s on the basis of the Texas City (TX, USA) disaster related to AN [38,39,40]. The detonations of AN resulted majorly from the influences of adjacent explosions. A sympathetic detonation (SD) is a phenomenon of coherent detonation induced by a donor detonation. The fact is that the SD is triggered passively by the concussion of explosion pieces or gigantic pressure impulse from the neighboring “explosion donor”. The detonation of high explosive occurs within a time of ms and the pressure is a value of GPa. When sympathetic detonation occurred, it took place a time scale less than 1 ms after impact from donor detonation. This means that the incident blast from a donor is extremely strong as the “acceptor explosive” detonates almost instantly. An investigation on the SD of AN regarding to the ammonium nitrate-fuel oil (AN-FO) was posed [38,39,40]. They indicated that an AN-FO acceptor against a 16-gage steel faced AN-FO donor, the SD could be ignited within a distance of 53 feet or equivalent to 16 charge diameters (i.e., the ratio of gap to diameter was equal to 16). However, with AN acceptor, the initiation occurred at approximately 5.5 charge diameters separation (i.e., the ratio of gap to diameter was equal to 5.5) [38,39,40]. The separation across which AN-FO and AN can be sympathetically detonated by an AN-FO donor was surprisingly huge, which had never been imagined. Their investigations showed that unexpectedly large safety distances inevitably hindered SD when the industry stored explosives or AN.

Until today, part of the occurred contingencies in the Tianjin incident were not declared. This study will stress the following major objectives/focuses: (1) how long will the first fire take from the NC container spontaneously ignite due to the autocatalytic decomposition, (2) whether the NC is more hazardous in closed or open state, (3) the assessment model for trinitrotoluene (TNT) equivalent in the explosion, (4) how the crater with a diameter of 97 m can be formed, and (5) to assess the fatality radius of the human body by relating the blast overpressure versus distance to provide safety measures, on-site rescues and emergency responses in an AN warehouse under fire scenario.

2. Materials and Methods

2.1. Samples

A sample of NC with a nitrogen content of 11.7% was supplied by a local company. NC was sealed in an explosion-proof container and stored below 20 °C.

2.2. Differential Scanning Calorimetry

Thermal instability of NC was screened using a Mettler DSC 3 system coupled with a STAR^e V15 control system [41]. Disposable aluminum crucible (ME-27731) was used to determine thermal curves under closed or open conditions. The exothermic onset temperature (T_onset) was chosen at the deflection point defined by ASTM E537 [42]. Scanning rate was selected to be 4 K min⁻¹ in temperature-programmed ramp. A typical isothermal aging curve with autocatalytic reaction can be verified by the characteristics of a long induction period, an acceleration period and a decay time. By obeying the ASTM E487 [43], the isothermal onset temperature can be defined as the maximum temperature at which a chemical compound or mixture may be held for a period under the conditions imposed on the test without exhibiting a measurable exothermic reaction. Taking a complete set of thermal curves using isothermal tests under several temperatures, the autocatalytic kinetics can be determined by following the ASTM E2070 [44]. The induction time (or termed as time-to-maximum-rate (TMR)) is the time an unstable compound or mixture held under isothermal conditions until it exhibits a distinct exothermic behavior [45]. TMR is a featured parameter to explain the induction time for the storage period of the NC from the time of arrival to that of the auto-ignition in conjunction with the first fire occurred.

2.3. Evaluation of Trinitrotoluene Equivalent

TNT equivalent is one of the cardinal physical quantities for evaluating the powers of explosions resulted from energetic explosives. Even if the detonation feasibility of AN was known not to be high, this catastrophic incident implied that this study on consequences must be performed for the purposes of safety measures and emergency responses. Traditionally, the overpressure of the explosion versus distance is evaluated, thus the simulated destructive influences on human bodies or demolition to buildings can be substantiated. For the modeling of detonations relating explosive chemicals, most of the traditional methodologies would like to use the TNT equivalent. One of the most significant viewpoints that must be weighed is to verify the safe distance for industry and community.

2.3.1. Trinitrotoluene Equivalent Model

A simplified approach applied the traditional TNT equivalent model from the conversion factor of explosive chemical into the effective TNT mass for assessing the power of shock wave resulted from explosion. It can be expressed as Equation (1) [46]:

$$\mathrm{TNT} \mathrm{equivalent} \mathrm{mass}=m_{\mathrm{TNT}}=m_{\mathrm{AN}}\times \frac{\Delta H_{\mathrm{AN}}}{\Delta H_{\mathrm{TNT}}}$$

(1)

where m_TNT is the equivalent mass of TNT, m_AN is the mass of AN, ΔH_TNT is the heat of decomposition of TNT, ΔH_AN is the heat of decomposition of AN, and ΔH_AN/ΔH_TNT is the equivalency efficiency of AN which links the realistic heat liberated in relation to TNT. The certainty for TNT equivalent can exhibit the heat generated to account for the blast powers correctly. AN possesses the m_TNT efficiency, which was generally adopted to be 0.35 by the heat of decomposition and Equation (1) [5,46].

2.3.2. Validation of Trinitrotoluene Equivalent

Damage–Distance Correlation

The endeavor to quantify the blast damage, distance, and m_TNT implicated an early study of blast power in the 1940s. The results of investigations into blast powers to buildings have been expressed in the following equation [47]:

$$R=\raisebox{1ex}{$A{m}_{\mathrm{TNT}}{}^{1/3}$}\!\left/ \!\raisebox{-1ex}{$\left[1+{\left(\raisebox{1ex}{$700$}\!\left/ \!\raisebox{-1ex}{$m_{\mathrm{TNT}}$}\right.\right)}^2\right]$}\right.{}^{1/6}$$

(2)

where R is the distance (in inch), A is a constant for a given “category” of destruction (in feet per pound^1/3), and m_TNT is the TNT equivalent (in pounds).

Empirical Formula of Scaling Law

Scaling laws for evaluating the crater sizes have been known worldwide on ground explosions driven by the chemical and nuclear tests till today [48]. These two kinds of explosions were demonstrated to be analogous in accounting the crater size as a power of m_TNT; however, which order was not necessarily equivalent to be the first. The scaling law was determined experimentally through practical tests of ground explosions. Scaling of crater size in relation to m_TNT was presented in the following expression:

$$D=c{m}_{\mathrm{TNT}}{}^n$$

(3)

where D is the crater diameter, n is the empirical index, and m_TNT is the TNT equivalent. Outcomes of reliable tests ranged from 1 kg to 5000 tons of TNT, however, which revealed that the constant n is not conventionally the same. In the earlier study, the diameter was proportional to the power of charge mass, ideally with a hypothetical index of 1/3 (i.e., the cube root of m_TNT) [49].

Table 2. Scaling laws for crater diameter related to TNT equivalent.

Author	Scaling Law	mTNT (kg)	Description
Vortman	D = 0.56 mTNT0.375	3.6, 29, 116, 454, 2720, 18,144	16 explosion tests on the surface of dry-lake at Nevada test site
Vortman	D = 0.40 mTNT0.408	232, 4564, 18,144, 90,718, 453,592	5 explosion tests at the Watching Hill site
Vortman	D = 0.32 mTNT0.426	3.63, 3.86, 13.83, 236, 250, 272, 4565, 4574, 4579, 18,144	30 explosion tests at the Downing Ford site
Adushkin and Khristoforov	D = 0.606 mTNT0.34	1000–5,000,000	54 explosion tests on surface
Ambrosini and Luccioni	D = 0.606 mTNT0.34	1, 2, 4, 7, 10, 250	6 explosion tests on surface
Ambrosini et al.	D = 0.80 mTNT1/3	1–10	30 explosion tests; Scaling law revised from tests by Kinney and Graham

summaries the scaling laws reported by Vortman [48] and compared those with explosion tests.

Inverse Analysis Based on Crater Size

On the basis of the abovementioned scaling law with a n of 1/3, any distance R from an explosive charge with mass m_TNT can be written by a scaled distance Z = R m_TNT^1/3. In the same approach, the scaled depth of crater can have the identical cube root related to m_TNT. Zhou et al. presented their calculations supported by eight-four results of the examined parameters of the craters created by TNT tests for inverse analysis [50].

2.4. Influences of Blast Overpressure on Buildings and Personnel

2.4.1. Shock Wave Blast Effects

Two of the most important factors, overpressure and impulse, representing the power of explosions are chiefly responsible for the destructive ability to humans, structures and environments. After the results obtained from this model, these data were then fitted to integrate equations showing the relationship among overpressure against distance or impulse in relation to distance, which are generally called “characteristic curves”. These characteristic curves associate mathematical equations or impulse expressions significantly with the m_TNT. In brief, characteristic curves are capable of displaying the explicit plots of overpressure in conjunction with distance. In this work, three methodologies of Baker et al. Sadovski et al. and Alonso et al. were employed to assess the results of harms to human bodies and the powers of demolition to structures [51,52,53].

2.4.2. Baker’s Methodology

Once m_TNT has been verified, the scaled distance Z_e can be assessed from the distance to the explosion point by scaling law, the Equation (4) can be applied for safe distance evaluation:

$$Ze=R/m_{\mathrm{TNT}}^{}{}^{1/3}$$

(4)

where Z_e is the scaled distance of TNT equivalent in mkg^1/3, R is the distance from the explosion origin in m, and m_TNT is the TNT equivalent in kg. Baker et al. utilized the following scaled distance of TNT equivalent to validate the relationship between overpressure and distance during detonation [51]. The formula for calculating the overpressure and safe distance of the shock wave generated by TNT explosion was expressed as follows:

$$\frac{\Delta P}{P}=\frac{1616\left[1+{\left(\frac{Z_e}{4.5}\right)}^2\right]}{\sqrt{1+{\left(\frac{Z_e}{0.048}\right)}^2}\sqrt{1+{\left(\frac{Z_e}{0.32}\right)}^2}\sqrt{1+{\left(\frac{Z_e}{1.35}\right)}^2}}$$

(5)

where ΔP is the overpressure of the shock wave in Pa and P is the ambient atmospheric pressure in Pa.

2.4.3. Sadovski’s Methodology

By assuming the pseudo-spherical origin of the explosion, the overpressure of shock wave analyzed from quite a number of tests was proposed by Sadovski [52]:

$$\Delta P=0.085\frac{\sqrt[3]{m_{\mathrm{TNT}}}}{\mathrm{R}}+0.3{\left(\frac{\sqrt[3]{m_{\mathrm{TNT}}}}{\mathrm{R}}\right)}^2+0.8{\left(\frac{\sqrt[3]{m_{\mathrm{TNT}}}}{\mathrm{R}}\right)}^3$$

(6)

where ΔP is the overpressure in MPa, R is the distance from explosion origin, and m_TNT is TNT equivalent in kg.

2.4.4. Alonso’s Methodology

By applying the methodology proposed by Alonso et al. [53], the overpressure can be confirmed by either one of the two regions, as expressed in Equation (7).

$$\Delta P=1.13\times {10}^6{Z}_e{}^{\left(-2.01\right)}$$

(7)

where ΔP is the overpressure in Pa, 1 ≤ Z_e ≤ 10 and 1.13 × 10⁶ ≥ ΔP ≥ 11,000.

For 10 ≤ Z_e ≤ 100 and 11,000 ≥ ΔP ≥ 400, the overpressure can be calculated by Equation (8) as follows:

$$\Delta P=1.83\times {10}^5{Z}_e{}^{\left(-1.16\right)}$$

(8)

Using the overpressure Equations (5)–(8), the characteristic curves can be validated and plotted. For the harm effects on human bodies stressed by Jeremić et al. [54] and Prugh [47] as well as other hurt limits underlined by Török et al. were presented [55]: 55 mbar for glasses rupture; 140 mbar for the beginning of lethality; 300 mbar for high lethality due to indirect effects of the explosion (impacted by flying objects or collapsing structure); 600 mbar for high lethality resulted from direct effects of overpressure.

3. Results and Discussion

3.1. Trinitrotoluene Equivalent

3.1.1. m_TNT Evaluation by Overpressure Effects

A distance of R = 300 m (948.25 ft) from the explosion crater, the construction of the traffic police detachment of the Tianjin Port Public Security Bureau was seriously demolished. By taking A = 9.5 on the basis of the constant proposed by Jarret [56], m_TNT in Equation (2) was calculated to be 1,112,104 lbs or 504 tons.

3.1.2. m_TNT Evaluation by Scaling Law

Based on practical tests, several constant C and index n in the Expression (3) were posed. Table 2 depicts the popular scaling law linked to the diameter and m_TNT [48,57,58,59].

Table 3. Crater diameters evaluated by the scaling laws and compared to explosion tests.

mTNT (kg)	D = 0.606 mTNT0.34 (m)	D = 0.80 mTNT1/3 (m)	D = 0.40 mTNT0.408 (m)
1	0.6	0.8	0.4
10	1.3	1.7	1.0
50	2.3	2.9	2.0
100	2.9	3.7	2.6
500	5.0	6.3	5.0
1000	6.3	8.0	6.7
5000	11.0	13.4	12.9
10,000	13.9	17.2	17.1
50,000	24.0	29.5	33.0
100,000	30.3	37.0	43.9
500,000	52.5	63.5	84.6
1,000,000	66.4	80.0	112.2
2,000,000	84.1	100.8	148.9
5,000,000	114.8	136.8	216.4
8,000,000	134.7	160.0	262.1
10,000,000	145.4	172.4	287.1

lists the crater diameters determined by the scaling laws, and explosion tests. Figure 2 shows the crater diameters in conjunction with the scaling laws, explosion tests and major incidents. Taking the photographs in Figure 1 into consideration, the diameter of the crater created by the explosion was 97 m. Therefore, the m_TNT of this incident by scaling law was verified to be 700 tons.

Comparisons: (1) 1921 Oppau AN explosion, m_TNT = 788 tons, D = 108 m (Pittman et al. 2014) [3], (2) 1964 TNT test at Watching Hill Site, m_TNT = 453.6 tons, D = 85.4 m (Vortman, 1968) [48], (3) 1993 Shenzhen AN explosion, m_TNT = 22.9 tons, D = 20 m (Jiang, 1998) [4], (4) 2001 Toulouse AN explosion, m_TNT = 157.5 tons, D = 60 m (Pittman et al. 2014) [3], (5) 2020 Beirut AN explosion, m_TNT = 950.3 tons, D = 112.9 m (6) 2015 Tianjin AN explosion, m_TNT = 700 tons, D = 97 m (this work).

3.1.3. m_TNT Evaluation by Inverse Analysis

The methodology of inverse analysis according to the relationship between m_TNT versus the scaled distance shown in corresponding figures was proposed by Zhou et al. [50]. All the parameters linked to m_TNT, including the diameter and depth of craters were plotted in the published journal [50]. By setting a vertical line at zero (i.e., ground explosion) in the Z vs. scaled depth of bursts (m/kg^1/3) plot, with an intersection at Z = D/m_TNT^1/3 = 1.2 and D = 97 m, the m_TNT can be calculated to be 528 tons for the major explosion.

3.2. Impact of Overpressure on Building and Human Body

In the following consequence analysis of the second explosion, by taking the average value of 577 tons from the three abovementioned m_TNT, the results between the resulting safety distances relative to overpressure using three different approaches were performed. In addition to Alonso’s methodology, by using the overpressure equation associated with the corresponding interval in Equations (4) or (5), the characteristic curves associated ΔP with R are determined. When m_TNT was chosen as 577 tons, the distance R using 300, 400, 500, 1000 and 3000 m being taken into the aforementioned groups of overpressure formulas, the overpressures were determined and presented in

Table 4. Evaluation of overpressure by three approaches.

Approach	ΔP (MPa) (300 m)	ΔP (MPa) (400 m)	ΔP (MPa) (500 m)	ΔP (MPa) (1000 m)	ΔP (MPa) (3000 m)
Baker’s	0.1111	0.0692	0.0425	0.0158	0.0047
Sadovski’s	0.0638	0.0379	0.0262	0.0096	0.0026
Alonso’s	0.0859	0.0482	0.0308	0.0076	0.0008

. The overpressure against distance can be decided and shown in Figure 3, where the consequences of the damaged structures were presented by solid points confirmed by the pictures in their corresponding positions. Consequently, side-on overpressure in relation to distance can be plotted on several concentric circle diagrams regarding to the satellite map presented in Figure 4. Five representative pictures of damaged structures for comparisons are depicted in Figure 5. In parallel, the effects of explosion impacts on the human bodies and on the structures can be further assessed through the overpressure effects proposed by Prugh [47].

One of the most intriguing parameters is to ascertain the separation distance, sometimes referred to as the safety distance. To reduce the deviations of m_TNT in evalution expressions, overpressure was evaluated by three aforementioned empirical methodologies. The results exhibited little differences in the safe distances acquired in the simulations among these methodologies. Characteristic curves shown in Figure 3 in accordance with each other were validated by the real demolished structures pictured in Figure 5. Expectantly, the safe distance decreased obviously versus overpressure in these three models, revealing that a safe enough distance can prevent or decrease the level of destruction on the communities, infrastructures, administrative regions and industries. It is clear that the m_TNT and safe distance both determine the destiny of risk and consequence of a detonation. In avoiding or diminishing the consequence of similar incident, it is cardinal to reevaluate the quantity of AN and safe distance for a warehouse or plant storing AN. Safe distances have to be considered from the following two points. The first is to prevent a domino effect or SD with enough separation for energetic chemicals stored in nearby zones. The second is that the m_TNT must be less than the allowable quantity required by international standards for production, operation, transportation, and storage [5]. It is distinct from Figure 3 that the overpressure resulting in death occurred at about 0.6 bar, which corresponds to the “fatality radius” intercepting the three characteristic curves at approximately 410 m from the explosion origin, illustrating that most victims were the firefighters involved in firefighting and some on-site employees of the Ruihai Company.

3.3. Autocatalytic Decomposition of Nitrocellulose

From the investigation report issued by the State Council, which clearly announced that the fire was directly caused by the auto-ignition of NC with a proof of video, literature related to the thermal hazards and chemical kinetics caused by the autocatalytic decomposition of NC was limited till now. Several studies demonstrated that the NC had an exothermic onset temperature at approximately 182 °C using differential scanning calorimetry (DSC) or TGA [17,18]. Figure 6 shows the thermal curve of exothermic reaction detected by a screening test using DSC. It is a regret that the traditionally non-isothermal method of DSC can provide the thermal curve, exothermic onset temperature, peak temperature, and enthalpy change. Exothermic phenomena of NC below 140 °C have never been observed by any calorimetry. To confirm the autocatalytic decomposition of NC, this study applied the isothermal tests regulated by both ASTM (American Society of Testing and Materials) E487 and E2046 to confirm the autocatalytic decomposition and detect the exothermic thermal curve associated with a distinctive time-to-maximum-rate [44,45]. Time-to-maximum-rate can be recognized as the synonym for the time to auto-ignition. Time-to-maximum-rate determined by DSC can be used to verify the storage time of NC in the accidental container from the time of arrival to auto-ignition. For DSC experiments the first temperature for isothermal test is suggested to be (T_onset − 10) °C by ASTM E 487, where T_onset is acquired from a non-isothermal screening test. All the standard isothermal tests must be implemented every minus 10 °C from the onset temperature until the exothermic signal cannot be detected. For simulating the spread of NC outside the package drum and in dry status, it was detected in an open crucible. At 150 °C the exothermic signal was weak and almost undetectable. However, it was astonishing that the closed crucible experiment could discriminate the exothermic curves as low as 90 °C. In this study, NC was first found to be more hazardous or unstable under a closed than an open state. A typical thermal curve related to the autocatalytic decomposition of NC at 120 °C is depicted in Figure 7. The time-to-maximum-rate was measured to be 1985.8 min at 90 °C. The maximum temperature on the day of the incident was reported to be 36 °C at Tianjin port; moreover, the inner space of the container might reach as high as 60 °C [1]. According to this, a time-to-maximum-rate with the value of 12,400 min or 8.6 days was determined at 60 °C in Figure 8. In 2015, the weather of Tianjin Port was sunny in most days from July to August, and containers held an internal average temperature of up to 60 °C. Under such circumstances the NC in the closed container decomposed autocatalytically. As to the containers cannot effectively dissipate heat, the partially superheated NC carried out faster exothermic decomposition to accompany the extreme temperature rise and gas production. This ruptured the container, which ignited itself automatically. “The results revealed that the NC was induced to reach the autocatalytic decomposition by the container with an internal temperature approximately 60 °C. The hot spot inside the NC bag gradually accumulated heat and accelerated the NC to the auto-ignition after 9 days”. Ignition occurred when the temperature of NC exceeded the AIT, which agreed with the recorded video. The burning substances spread to neighboring containers containing flammable or combustible goods, then extended to more containers loaded with AN. Under the strong influence of the surrounding fire, the AN containers exploded in the long run. The isothermal TMR of autocatalytic decompositions under open and closed pans were listed in

Table 5. Isothermal TMR of nitrocellulose in DSC open pan tests.

Temperature (°C)	Mass (mg)	TMR (min)
180	1.2	25.4
175	2.2	36.4
170	2.2	65.5
165	4.0	120.3
160	4.0	195.3

and

Table 6. Isothermal TMR of nitrocellulose in DSC closed pan tests.

Temperature (°C)	Mass (mg)	TMR (min)
170	2.2	46.1
160	1.5	47.1
150	1.5	65.0
140	1.5	78.9
130	1.5	106.5
120	1.5	132.8
110	1.5	357.3
100	2.2	859.0
90	1.5	1985.8

, respectively. Based on the chemical kinetics for autocatalytic decomposition of NC reported by Hai et al. [24], a simulation curve was depicted in Figure 8; it demonstrated that the time took more than 60 days for NC to auto-ignite at 60 °C. The induction was too long to be acceptable and cannot coincide with that of the incident. The Ruihai company received the business approval on 24 June, and it was recognized that the time was less than 20 days after the arrival of NC to auto-ignition.

3.4. Incidents of Sympathetic Detonations

SD is typically induced by the propagation of detonation between two explosive chemicals. The neighboring acceptors adjacent to a detonating charge can be collided by the flying fragments and be stimulated by the shock wave generated from donor detonation. Experimental results exhibited that the detonation speed was extremely dependent on the size of the donor. A large number of processes leading to SD include initiation of explosion in the donor charge, an increase in pressure, the scattering of fragments of an exploding charge, and penetration of the high-speed projectiles into the target. Van Dolah conducted a large-scale test program using AN [38]. By analyzing the induced SD declared that from the initiation origin, it must span across 47 ft. to detonate acceptors sympathetically. The description of the AN explosion in Oppau at 1921, in accordance with the witnesses there, had two successive explosions, the first one being faint or unclear and the second one being fatal [3]. Another explosion of AN in West Fertilizer Company, an explosion started as a fire, which resisted for only 20 min then detonated after the 911 call. On 4 May 1988, a fire began in a large ammonium perchlorate (AP) warehouse located in Henderson, Nevada. This specific SD recognized as a specific incident was elucidated by Mniszewski [60]. Two large explosions occurred during the fire, each on the order of several hundred tons of TNT equivalent. A time interval of this SD propagation was first recorded as short as 4 min. Marlair et al. [61] summarized about nine AN fire scenarios that led to the subsequent explosions. In 2014 at Charleville, a transport accident of truck rollover was followed by an “explosions-following-a-fire-scenario” [61]. After 75 min passed from the time of the fire, the major explosion followed the first explosion about 1–2 min. Some articles examined the likelihood of an explosion of AN when exposed to fire scenarios, the detailed mechanism of an SD of AN under fire was obscure. Babrauskas [30] identified 58 separate AN incidents from 1920 to 2014, which were caused by an uncontrollable fire; from these events only 17 explosions occurred. It is identified that, for AN explosion in storage or transportation, an uncontrollable fire is the essential scenario but not decisive. It is conclusive that the catastrophes caused by AN detonations can be prevented or mitigated by extinguishing the fire before it being uncontrollable.

Table 7. Major incidents of SD with the “two-successive-explosions-following-a-fire” scenario.

Incident/Year	Chemical	Quantity (ton)	Ignition (Time)	First Explosion (Time)	Major Explosion (Time)
Oppau/1921	(NH4NO3)(NH4)2SO4	5000	Blasting at 07:00	07:31	07:32
Nevada/1988	NH4ClO4	Several hundred tons	Fire at 11:15	11:53	11:57
Shenzhen/1993	AN	65	Fire at 13:10	13:26	14:27
Queensland/2014	AN	53	Fire by truck rollover at 09:00	10:10	10:12
Tianjin/2015	AN	430	Fire at 22:51:46	23:34:06	23:34:37

lists the major incidents of SD with the “two-successive-explosions-following-a-fire” scenario [30,60,61].

3.5. Scenario of Two-Successive-SD-Following-a-Fire in Tianjin Incident

Whether AN will detonate or not was decided significantly by the crucial conditions such as the dimension of the particle, porousness of AN, confinement, fire, charge quantity and bulk density. A project for the study on SD of AN was executed by the U.S. Bureau of Mines [38]. Application of 5400 lb ANFO with a donor size 60 inches and metal skin, a separation distance of 153 ft (43.3 m) for avoiding the sympathetic detonation was obtained [38,39]. Additionally, safety distances were confirmed to increase due to bin preheating for simulation of fire case. A 20′GP container with 5.69 m × 2.13 m × 2.18 m internal dimensions or volume of 33.2 m³ can load about 10,000 kg of NC or 50,000 kg of AN. According to analysis using a container with a size of 80 inches (2.032 m), the SD must span across 232 ft (70.7 m), propagating the blast impact across all containers for the second disastrous explosion in Tianjin incident. For the incident, from the aforementioned scaling law, the second SD had the m_TNT of 700 tons to create the crater with a diameter of 97 m. A m_TNT of 700 tons was equal to approximately 1999 tons of AN (loaded in 39 containers). This fact showed that all the stacked AN in containers could explode by the phenomenon of SD due to the insufficient separation distance. It was deduced that the propagation speed of SD in the two explosions was almost 3000 ms⁻¹, an SD initiated inside a container needed near 0.006 s. At this rate, it traversed the second SD spending about 0.3 s (39 containers times 0.006 s for each container).

4. Lessons Learned

To prevent a similar major AN incident to the Tianjin disaster, it is of great importance to propose the lessons learned for the public, industry, stakeholder, regulatory body and government. This incident encompassed the auto-ignition by autocatalytic decomposition of NC in container at summer, SD of AN with an average quantity of 1649 tons. Some concluded lessons are suggested to be learned:

• From the amount of fatality and injury, this is the most serious explosion of AN in the world since 1947 and before the incident at the Beirut Port occurred in 2020.
• Most of the victims and injured individuals were firefighters, therefore the guides for fire fighting in an industrial plant or warehouse storing AN must be carefully planned before their actions for emergency responses [62].
• The Ruihai company illegally stored too much AN and other dangerous chemicals.
• Reactive substances, NC and AN were stacked together to create the difficulties for fire-fighting and the risks of fire hazards.
• NC can be ignited by autocatalytic decomposition in containers during summertime due to poor heat removal.
• The hazards of autocatalytic decomposition of NC were proven to be more severe in closed than open status.
• NC is determined be ignited after about 9 days by autocatalytic decomposition in a container sustained at 60 °C.
• AN in a confined container can explode under external fire and propagate severe SD.
• A safety distance to prevent the SD caused by AN in a 20′GP container was determined to be 97 m in Dolah’s studies, indicating all the stacked AN containers will explode within a few seconds by SD if one of them ignites the first explosion.
• The m_TNT in conjunction with the second explosion was averaged to be 577 tons using scaling law, damage-distance correlation and inverse analysis.
• NC and AN were reported to be the very materials to ignite the terrible first fire and disastrous explosion, respectively.
• The incident was a typical event of “two-successive-explosions-following-a-fire”.

5. Conclusions

AN has attracted attention from authorities over recent years due to various accidents. A study of the calamitous incident of Tianjin Port was implemented and discussed. The first fire initiated by NC under autocatalytic decomposition in a confined container was validated through ASTM standards and isothermal tests of DSC. The time to ignition of NC was validated to take 9 days in a container sustained at 60 °C. Suffering from the spreading fire and the blackbody radiations, the super-heated NC or AN underwent the first explosion under a thermal runaway. According to the formula relating to the influences of impulsing wave on structures, inverse analysis and the scaling law related to m_TNT, the average m_TNT was quantified to be 577 tons relative to AN of 1649 tons. Three different methodologies have been implemented to validate the aftereffects of the explosion. In this disastrous event, the destroyed buildings were essentially coincident with those photos posed, that moreover demonstrated the reliability of the power of the second explosion speculated. From the blast effects and overpressure to human body, a “fatality radius” was authenticated to be 410 m from the origin of explosion.

This incident exactly revealed a distinct scenario of “a fire following two successive explosion”. Two explosions were corroborated to be sympathetic detonation; the first was not strong, but the second was crucial because more AN containers were stacked to be triggered and then suffered the SD. The explosion of a small amount of AN in a confined space may ignite the explosion of large quantity of AN via the effects of SD. The Ruihai International Logistics Co., Ltd. stored too huge quantity of labile NC and without corrective actions to prevent the SD of AN when catching fire. More through studies have to be conducted before the characteristics of fire/explosion of AN can be completely elucidated. As a guide to preventing potential catastrophe in the near future, this holistic case study discloses the importance of strengthening the existed regulations and promoting the risk awareness to store and operate AN safely by reducing the storage quantity.