Factors affecting the combustion toxicity of polymeric materials

Abstract

Fire gas toxicity is an essential component of any fire hazard analysis. However, fire toxicity, like flammability, is both scenario and material dependent. A number of different methods exist to assess the fire toxicity, but many of them fail to relate this to a particular fire scenario. Sample thickness alone, in a closed box test such as the NBS Smoke Chamber, is shown to change the fire scenario from well-ventilated to under-ventilated. Data from two flow-through tests, the static tube furnace (NF X 70-100) and the steady state tube furnace (the Purser furnace, BS 7990 and ISO TS 19700) show that there are different patterns of behaviour for different polymers (LDPE, polystyrene, rigid PVC and Nylon 6.6). The predicted toxicities show variation of up to two orders of magnitude with change in fire scenario. They also show change of at least one order of magnitude for different materials in the same fire scenario. Finally, they show that in many cases CO, which is often assumed to be the most, or even the only toxicologically significant fire gas, is of less importance than either HCl, or HCN, when present, and in some cases less important than organo-irritants. Nylon 6.6 shows the highest predicted toxicity, the greatest scenario dependence, and the least sensitivity to different apparatuses, while polystyrene shows the highest sensitivity to the different apparatuses, but the lowest to different fire scenarios. PVC shows high toxicity, mostly due to HCl in the fire effluent, under all fire conditions, and LDPE shows a more progressive increase in toxicity from well-ventilated flaming to both smouldering and under-ventilated flaming.

Introduction

The essential reason for incorporating fire retardants into materials is to reduce the hazard to life from fire. Fire retardant strategies aim to reduce the ignitability or the rate of heat of heat release of burning items. However, fire hazard is a combination of both the flammability and the fire gas toxicity, and since most UK fire deaths [1] are attributed to the effects of toxic gas and smoke it is pertinent to identify the factors affecting the toxicity of burning polymers. In addition, most unwanted fires involve combinations of materials, and thus fire retarded materials will still be involved in fires despite their lower flammability. There are a large number of different methods for determination of the toxic potency of fire effluents from materials or products. These different methods yield apparently inconsistent data because they represent different fire scenarios; measure product yields either as a function of material flammability or independent of it; base the toxicity assessment on the concentrations of different species; or use animal exposure to generate an overall estimate of toxic potency without knowledge of the relative contributions of the chemical species.

Fire effluents present two hazards to human life, as toxicants causing collapse and death directly, or as incapacitating irritants, impairing the function of the lungs and eyes, preventing escape. The effects of the two main toxicants are fairly well understood, carbon monoxide (CO) causes death by binding strongly to haemoglobin, preventing the transport of oxygen from the lungs to the body, while both CO and hydrogen cyanide (HCN) inhibit metal containing enzymes, including cytochromoxidase, which provides energy within cells by reaction with oxygen. Much less well understood are the two effects of acid gases, irritant organic species, and particulates. First, certain species such as isocyanates, acid gases, aldehydes, styrene and phenols stimulate pain receptors in the eyes and upper respiratory tract, resulting in inflammation and fluid release (acute bronchitis) when nerves respond to acidic and organic irritant gases, inhibiting breathing and causing the respiration rate to fall to 10% of its normal value. Where guinea pigs were forced to keep moving on a wheel when they were exposed to these concentrations, this resulted in death, rather than mere respiratory suppression [2]. Second, deep in the lung, the airways (bronchioles) become very finely divided, ultimately terminated with alveolae, where oxygen and CO₂ pass through the blood-gas barrier. The general effect of organic molecules and particulates here is to cause fluid release and inflammation, preventing gas exchange. Thus the body's response to fire gas is to shut down respiration as far as possible to minimise further toxic effects, but in so doing, deprives the cells of sufficient oxygen to allow escape.

The criteria used for assessing smoke toxicity from chemical analysis of fire effluents and how it should be measured have been defined [3]. The methodology is based on the results of experiments using exposure of rats to the common fire gases. The general approach is to assume additive behaviour of individual toxicants, and to express the concentration of each as its fraction of the lethal concentration for 50% of the population (LC₅₀). Thus an FED equal to one indicates that the sum of concentrations of individual species will be lethal to 50% of the population over a 30-min exposure.

$$\text{FED}=\frac{m\left[\text{CO}\right]}{\left[{\text{CO}}_2\right]-b}+\frac{21-\left[{\text{O}}_2\right]}{21-{\text{LC}}_{50,{\text{O}}_2}}+\frac{\left[\text{HCN}\right]}{{\text{LC}}_{50,\text{HCN}}}+\frac{\left[\text{HCl}\right]}{{\text{LC}}_{50,\text{HCl}}}+\frac{\left[\text{HBr}\right]}{{\text{LC}}_{50,\text{HBr}}}+\dots$$

Eq. (1), referred to as the N-Gas model, uses this approach. However, in the first term the effect of the CO is enhanced by the increase in respiration rate caused by high concentrations of CO₂ (expressed as a step function with one value of constants m and b for CO₂ concentrations below 5% and another for those above 5%). Alternatively the toxicity can be expressed as an LC₅₀, which in the case of a burning polymeric material is the specimen mass M which would yield an FED equal to one within a volume of 1 m³ on burning. The relation to the FED from the N-Gas model [3] is given in Eq. (2).

$${\text{LC}}_{50}=\frac{M}{\text{FED}\times V}$$

where V is the total volume of air in m³ at STP. The accuracy of LC₅₀ values determined in this manner is quoted as ±30% if the concentrations of all the contributing toxicants are measured and included [3].

Although all fires may be regarded as unique, burning behaviour and toxic product yields depends most strongly on a few of factors. Amongst them material composition, temperature and oxygen concentration are normally the most important. The generalised stages in the development of a fire have been recognised and are used to classify fire growth into a number of stages, from smouldering combustion and early well-ventilated flaming, through to fully developed under-ventilated flaming [4]. The stages of a fire, from non-flaming, to well-ventilated flaming, and finally to under-ventilated flaming, have been classified by ISO (Table 1), in terms of heat flux, temperature, oxygen concentration (to the fire, and in the fire effluent), and CO₂ to CO ratio, equivalence ratio Φ (the actual fuel-to-air ratio divided by the stoichiometric fuel-to-air ratio) and combustion efficiency (the % conversion of fuel to fully oxygenated products, such as CO₂ and H₂O.

For most materials the yields of toxic species have been shown to depend critically on the fire conditions. Fig. 1 illustrates the generalised change in toxic product yields during the growth of a fire from non-flaming through well-ventilated flaming to restricted ventilation. Although the toxic product yields are often highest for non-flaming combustion, the rates of burning and the rate of fire growth are much slower, so under-ventilated flaming is generally considered the most toxic fire stage.

The influence on fire conditions on the toxic product yield can be illustrated by consideration of the formation of carbon monoxide (CO) and higher hydrocarbons in methane flames [5]. The free radical processes typical of flaming combustion lead to very rapid reactions, once a critical free radical concentration has been reached. Since the number of radicals is much smaller than the number of molecules, any reduction in the radical concentration is likely to lead to incomplete combustion. The methyl radical CH₃ is formed by attack of an existing radical on methane.

CH₄ + OH → CH₃ + H₂O

CH₄ + H → CH₃ + H₂

CH₄ + O → CH₃ + OH

In the presence of sufficient oxygen, the methyl radical CH₃ reacts with oxygen to form methanal (formaldehyde).

CH₃ + O₂ → HCHO + OH

Since very little methanal is observed this must be followed by a rapid removal step such as

HCHO + OH → HCO + H₂O

and then

HCO + OH → CO + H₂O

In reduced oxygen environments two methyl radicals can combine, forming ethane, which may then be oxidised to acetaldehyde or dehydrogenated to ethane, then ethyne, benzene, higher aromatics and ultimately a carbonaceous soot particle. In higher oxygen environments, carbon monoxide is consumed is by reaction with a further OH radical:

CO + OH → CO₂ + H

For ignition and fire growth to occur, there must be an increase in the number of free radicals, involving chain branching steps. This will generate the species responsible for the initial attack on methane.

H + O₂ → OH + O

O + H₂ → OH + H

H₂O + O → OH + OH

Most of the OH radicals are produced by the reaction:

H + H₂O → H₂ + OH

Since OH radicals are relatively scarce, this prevents oxidation of CO until higher in the flame. Thus, crucially, if the H radicals cannot diffuse back to the lower part of the flame, they will not be able to generate enough hydroxyl radicals (OH) to convert CO to CO₂. Thus although CO always results from incomplete combustion, this can arise from insufficient heat (e.g. during smouldering, which is dominated by reactions at the fuel surface, rather than gas phase free radical processes); quenching of the radical flame reactions (e.g. when halogens are present in the flame forming free radicals which are stable enough to leave the flame zone without further reaction, or excessive ventilation cools the flame); the presence of stable molecules which do not succumb to attack by free radicals, such as aromatics which survive longer in the flame zone, giving high CO yields in well-ventilated conditions, but lower than expected yields in under-ventilated conditions [6]; and insufficient oxygen reducing the availability of OH radicals for the CO oxidation stage (e.g. in under-ventilated fires, large radiant heat fluxes pyrolyse the fuel even though there is not enough oxygen to complete the reaction).

In large fires, unless air has direct access to the base of the flame, the centre of the flame is always under-ventilated, and the flame height at which there is an opening at the tip of the flame and smoke starts to appear (known as the smoke point height), is dependant on the nature of the fuel [7], and there will always be a significant yield of CO, hydrocarbons and smoke.

Data from large-scale fires [8], [9] shows much higher levels of the two of the major toxicants (CO and HCN) under conditions of reduced ventilation. It is therefore essential to the assessment of toxic hazard from fire that these different fire stages can be adequately replicated, and preferably the individual fire stages treated separately. The drive for internationally harmonised methods for assessment of combustion toxicity, through adoption of international standards, such as those of ISO, provides the framework for meaningful and appropriate use of toxic potency data in the assessment of fire hazard. As structures and means of transportation become larger and more complex, there is movement away from the more traditional methods of ensuring fire safety by prescriptive codes, towards fire risk assessments and engineering solutions. Reliable rate of heat release, fire effluent toxicity and smoke generation data are all essential elements of such an assessment.

Certain bench-scale determinations of toxic product yield apply only to well-ventilated burning, such as the cone calorimeter [10] (although some materials, such as polystyrene, have reached the smoke point height at the scale used) or ASTM E1678 [11]. Others use a closed box apparatus, where a sample is combusted in a fixed volume of air, with natural or forced air circulation, and the mass of sample burnt can dictate whether the fire condition is well-ventilated or under-ventilated. To illustrate the effect of sample thickness, which changes the fuel-air ratio from well-ventilated to under-ventilated in a closed cabinet test, the relationship established by Tewarson [12], has been used to relate the yields to the equivalence ratio Φ. The yields of carbon dioxide, carbon monoxide and hydrocarbons (HCs) for polyethylene specimens of different thicknesses have been predicted for burning in a fixed volume of 0.51 m³, the volume of the NBS Smoke Chamber. The yields of CO and hydrocarbons were calculated by Eqs. (1) and (2).

$${\text{Y}}_{\text{CO}}=0.024\left(1+\frac{10}{{\mathrm{e}}^{2.5{\varphi }^{-2.8}}}\right)$$

$${\text{Y}}_{\text{HC}}=0.007\left(1+\frac{220}{{\mathrm{e}}^{2.5{\varphi }^{-2.5}}}\right)$$

The yields were obtained by numerical integration over the range of Φ values, starting from well-ventilated, as the oxygen is consumed by the different masses of burning polyethylene, using a 1-mm step size. The yields have been converted into predicted concentrations if the fire effluent from the NBS Smoke Chamber were diluted from 0.51 m³ into a volume of 2.0 m³ (since FED values should be close to 1 when determined, and because the lack of oxygen in the undiluted fire gas would already render it lethal). The gas concentrations were input into equations to predict the lethality (to rats) as described in ISO 13344 [3], with the addition of a value [13] of 10 mg l⁻¹ to estimate the concentration of hydrocarbon gases causing incapacitation. This allows the LC₅₀ (the mass of PE burnt per cubic metre to create a lethal fire effluent) of the polyethylene to be predicted. This quantity might initially be expected to be a constant for a particular polymer, but in fact it is highly dependent on the fire condition. For closed box tests, the fire condition is mainly a function of the amount of fuel, in this case indicated by specimen thickness. Table 2 shows how the LC₅₀ varies as a function of Φ from 48.5 g m⁻³ at a thickness of 5 mm to 27.3 at a thickness of 15 mm. Thus the predicted toxicity, per gram, almost doubles with increase in sample thickness. The ISO standard [14] for this method states that the results are only valid for the thickness tested. This shows that the LC₅₀ is not a material property, but is highly dependant on the fire conditions; in closed cabinet tests, with materials of unknown composition, it is not easy to define the fire condition.

The contribution of CO and hydrocarbons to the fractional effective dose (FED) has been estimated. Fig. 2 shows the contributions to the predicted FED from different thicknesses of polyethylene in the Smoke Chamber. This shows a sharp increase in the CO and HC yield with thickness of polyethylene sample as the fire condition changes from well-ventilated to under-ventilated flaming.

When the FED reaches a value of one, the gas concentration would be lethal to 50% of the population in a 30-min exposure. The figure shows a disproportionate increase in the predicted toxicity per gram of polyethylene when diluted to an arbitrary 2.0 m³ as the thickness of the sample increases from 1 to 15 mm and illustrates that the effect of oxygen depletion on the CO and hydrocarbon yields. The theoretical example has been limited to CO, CO₂, and hydrocarbons for the NBS type, closed box test, but shows the general effect of test apparatus and operating parameters on fire conditions and clearly a similar effect also applies to other closed cabinet tests and to some flow-through tests.

One example of a static flow-through method, where the fire stage may be unknown is the French Railway test (NF X 70-100). This is a small-scale static tube furnace (Fig. 3) with a fixed ventilation rate, using furnace temperatures of 400, 600 and 800 °C to represent smouldering, well-ventilated and under-ventilated conditions. At a temperature of 600 °C, the rate of burning may be fairly steady, and well-ventilated, whereas at 800 °C, the fire condition may be closer to under-ventilated as the rate of pyrolysis exceeds the air supply rate. Other tube methods such as the DIN 53436 and the UK steady state tube furnace, BS 7990 and ISO TS 19700 (The Purser furnace) allow the possibility of controlling the fire conditions during burning. These methods force combustion by feeding the sample into a furnace of increasing heat flux at a fixed rate, thus replicating each fire stage by steady state burning. Ultimately the value of the bench-scale toxicity assessment is dependent on its ability to predict large-scale burning behaviour, and therefore validation must involve comparison with large-scale test data. Unfortunately most large-scale test data have been obtained under well-ventilated conditions, and when under-ventilated fire scenarios, such as the ISO 9705 Room test, are used the change of sample mass and the air flow to the fire during the test is not generally known.