MAGNETEK: A Case Study in the Daubert Challenge

Reading List

Babrauskas, Vytenis, Ph.D. “Truck Insurance v. MagneTek: Lessons to be Learned Concerning Presentation of Scientific Information.” Fire & Arson Investigator 55(2):9-10, October, 2004.*

Babrauskas, Vytenis, Ph.D. “Pyrophoric Carbon…The Jury is Still Out.” Fire & Arson Investigator 51:2, 12-14, Jan. 2001.*

Babrauskas, Vytenis, Ph.D. “Ignition of Wood: A Review of the State of the Art.” pp. 71-88 Interflam 2001, Interscience Communications Ltd., London 2001.*

Browne, F.L. “Theories of Combustion of Wood and its Control, A Survey of the Literature.” Report No. 2136. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory. 1958*

Burnette, Guy E. Jr. “Magnetek: New Case Applying the Daubert Standard” Internal Report, Guy E. Burnette, Jr. P.A., 2004.*

Burnette, Guy E. Jr. “Daubert Revisited: Magnetek and the Death of Pyrolysis.” Fire & Arson Investigator 55(1):47-49, July, 2004.*

Burnette, Guy E. Jr. “Daubert Revisited: Bitler v. A.O Smith Corp.” Internal Report, Guy E. Burnette, Jr. P.A., 2005.*

Cuzzillo, B.R.; Pagni, P.J. “Myth of Pyrophoric Carbon.” Proceedings. Sixth (6th) International Symposium July 5-9, 1999. International Association for Fire Safety Science.*

Fireman’s Fund Insurance Company v. Canon U.S.A, Inc., 8th Cir., No. 03-3836*

LeVan, S.L. “Thermal Degradation.” Concise Encyclopedia of Wood & Wood-Based Materials, 1st Edition, pp. 271-273, 1989.

White, Robert H., and Mark A. Dietenberger. “Chapter 17 - Fire Safety” Forest Products Laboratory. Wood handbook—Wood as an engineering material. Gen. Tech. Rep. FPL–GTR–113. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory, 1999.*

White, Robert H., and Mark A. Dietenberger. “Wood Products: Thermal Degradation and Fire” Encyclopedia of Materials: Science and Technology ISBN: 0-08-0431526 pp. 9712-9716 5

“Ignition and Charring Temperatures of Wood.” Report No. 1464. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory.*

NOTE: References identified with a (*) are included in the MagneTek: A Case Study in the Daubert Challenge participant handout.

IAAI Fire Investigator Distance Learning Project JCJ 2-11-2005

GUY E. BURNETTE, JR., P.A. ATTORNEYS AT LAW Guy E. Burnette, Jr. Lucy P. Hassler, Paralegal Marc A. Peoples Tami M. McClane Paralegal Jeanie L. Martin, Office Manager

MAGNETEK: NEW CASE APPLYING THE DAUBERT STANDARD

The decision of the United States Supreme Court in Daubert v. Merrell Dow Pharmaceuticals, Inc., 590 U.S. 579, 125 L. ED. 2d 469, 113 S. CT. 2786 (1993) forever changed the field of expert testimony in federal court cases. It has subsequently been adopted in a number of state court jurisdictions across the country as the basis for evaluating and admitting expert testimony. In the years since the Daubert decision, expert witnesses in virtually all fields have been subjected to a new level of scrutiny by trial judges acting as the “gatekeepers” of expert testimony in court proceedings. In fire litigation cases, the effect of this decision has fundamentally altered the process of proving the origin and cause of a fire in every type of case, from a product liability action to an arson defense.

A recent decision of the United States Court of Appeals for the Tenth Circuit has underscored the importance of properly evaluating and presenting expert testimony in fire litigation cases. Truck Insurance Exchange v. MagneTek, Inc., 2004 U.S. App. LEXIS 3557 (February 25, 2004) not only demonstrated the application of the Daubert standard for the admissibility of expert testimony, but effectively undermined a long- standing theory of fire science which has been used in a number of previous cases to prove the cause of a fire.

The MagneTek case involved a subrogation action filed by Truck Insurance Exchange against the manufacturer of a fluorescent light ballast alleged to have caused a fire which destroyed a restaurant in Lakewood, Colorado. When the fire department first responded to the alarm, they found heavy smoke in the restaurant but no open flames. The fire subsequently broke through the kitchen floor in the restaurant from the ceiling of a storage room in the basement. Before the fire could be controlled and extinguished, it destroyed the restaurant and caused damages in excess of $1.5 million.

The fire was investigated by the local Fire Protection District and a private fire investigation firm hired by the insurer. Following a thorough investigation of the fire scene, it was determined that the fire had started in a space between the basement storage room ceiling and the kitchen floor. In the basement, the investigators found the remains of a fluorescent light fixture that had been mounted to the ceiling of the storage room.

December 6, 2004

They determined the light fixture had been located in the area of origin of the fire and concluded that the fire had been caused by an apparent long-term failure of the ballast in the light fixture.

The remains of the fluorescent light fixture were examined by the investigators and an electrical engineer. They determined the ballast had been manufactured by MagneTek. They observed oxidation patterns on the light fixture indicating an internal failure, along with discoloration of the heating coils of the ballast suggesting it had shorted to cause overheating which resulted in the fire.

The ballast contained a thermal protector designed to shut off power to the fixture when the internal temperature exceeded 232 degrees Fahrenheit. The thermal protector in the ballast was tested and appeared to function properly even after the fire. However, the investigators remained convinced that the ballast had somehow failed, overheated and started the fire.

Tests were conducted with similar ballasts manufactured by MagneTek which showed that at least one of the exemplar ballasts when shorted would not cut off power to the fixture until the internal temperatures had reached at least 340 degrees Fahrenheit and would continue to provide power to the fixture even when the ballast maintained constant temperatures of 300 degrees or more. The investigators theorized that the heat from the ballast had caused pyrolysis to create “pyrophoric carbon” in the adjacent wood structure of the ceiling over a prolonged time, which would be capable of ignition at temperatures substantially below the normal range of 400 degrees Fahrenheit or more for the ignition of wood. The theory of pyrolysis and the formation of pyrophoric carbon has been the subject of a number of studies, reviews and articles by fire investigators and fire scientists. It has been cited as the cause of a number of fires having no other apparent explanation, often linked to overheated wires in structures within walls, ceilings and floor areas. As the MagneTek court would note, however, the validity of this theory has been discussed and debated by fire investigators and fire scientists for a number of years.

The investigators in this case acknowledged that electrical wiring ran through the ceiling area of the storage room near the fluorescent light fixture, but discounted the possibility of a failure in the electrical wiring. They reported finding no evidence of arcing or shorting in the electrical wiring, although the fire at the restaurant resulted in the destruction of most of the electrical wiring evidence in the area. Because they concluded the fire had originated in the immediate area of the light fixture, they concluded the only source of ignition for the fire was the light fixture and its ballast. The theory of the formation of pyrophoric carbon was the foundation of the plaintiff’s case against MagneTek. The ballast in the light fixture showed no signs of failure in the thermal protector, which would have limited the heat generated by the ballast to about 232 degrees Fahrenheit. Even with the exemplar ballast whose thermal protector failed to perform as

December 6, 2004 it had been designed, the temperatures generated did not exceed 340 degrees Fahrenheit. The investigators admitted the ignition temperature of wood is typically at least 400 degrees Fahrenheit and the temperatures from the ballast alone would not have been sufficient to cause ignition. Their theory that the ballast had caused the fire depended upon the concept of pyrophoric carbon to allow ignition to occur at a lower temperature within the range of the temperatures generated by the ballast.

Following discovery in the case, MagneTek filed a “Daubert Motion” to exclude the testimony of the experts that the ballast had caused the fire. MagneTek asserted that the theory of pyrolysis was not sufficiently reliable and scientifically verifiable to be offered by the experts in support of their conclusion for the cause of the fire. It was a challenge to the “reliability” of the experts’ theory which required a consideration of whether the reasoning and methodology underlying the testimony was scientifically valid as mandated by Daubert and Rule 702 of the Federal Rules of Evidence.

The Supreme Court in Daubert had outlined a number of factors that, while not an exclusive list of considerations for a trial court, should be examined in making the determination of reliability. Those factors include: (1) whether the opinion has been subjected to testing or is susceptible of such testing; (2) whether the opinion has been subjected to publication and peer review; (3) whether the methodology used has standards controlling its use and a known rate of error; and (4) whether the theory has been accepted in the scientific community. Daubert at 590.

In proving the scientific validity of an expert’s reasoning and methodology, the court noted that “the plaintiff need not prove that the expert is undisputably correct or that the expert’s theory is ‘generally accepted’ in the scientific community. Instead, the plaintiff must show that the method employed by the expert in reaching the conclusion is scientifically sound and that the opinion is based on facts which sufficiently satisfy Rule 702's reliability requirements”.

The Tenth Circuit applied the standard of appellate review for a trial court’s ruling on a Daubert issue: the showing of an “abuse of discretion” demonstrating that the appellate court has “a definite and firm conviction that the lower court made a clear error of judgment or exceeded the bounds of permissible choice in the circumstances.” United States v. Ortiz, 804 F.2d 1161 (10th Cir. 1986). The court then looked to the ruling of the trial judge finding that the testimony of the experts failed to satisfy the reliability standard under Rule 702 and the Daubert decision. The electrical engineer testifying on behalf of Truck Insurance Exchange was a highly credentialed expert with an advanced degree in physics from Oxford University and over twenty (20) years of experience in the study of fire and explosion incidents. Both the trial court and the appellate court observed that this expert was unquestionably qualified to testify as an expert witness under Rule 702. However, his hypothesis of pyrophoric carbon as the cause of ignition of the wood in the area surrounding the light fixture could not be considered a reliable basis for the admission

December 6, 2004 of his expert testimony on the cause of the fire.

The reliability standard under Daubert applies to both the reliability of the theory itself and the reliability of its application to the facts of the case. The court focused upon the first component of this reliability criteria. In support of the theory of its experts, the insurer had introduced into evidence three (3) publications on the theory of pyrophoric carbon. Those articles were written by some of the most respected fire scientists in the world, but those articles and case studies acknowledged that the process of pyrophoric carbonization occurred over an undefined period of time described as “a period of years” or “a very long time” with no specific parameters for the timing and sequence of events. One of those articles acknowledged that there are “a number of things not known about the process” and that “it may be many decades before it will be solved by sufficiently improving theory.” The article concluded by stating that “the phenomenon of long-term, low-temperature ignition of wood has neither been proven nor successfully disproven at this time.” The plaintiff’s engineer had stated in his deposition that the process “depends on a lot of factors, as yet quantitatively unidentified.” He went on to testify that “you would have to have a good theory of pyrophoric carbon and formation and the chemical kinetics of that; and there isn’t one. . .” The other experts testifying for the plaintiff in this case based their theory of pyrophoric carbon on their experience in the investigation of fires without any reference to a specific scientific basis for that theory. The court noted that under NFPA 921 an investigator offering the hypothesis of an appliance fire must first determine the ignition temperature of the available fuel in the area and then must determine the ability of the appliance or device to generate temperatures at or above the ignition temperature of the fuel. In that regard, the experts failed on both counts. The ignition temperature of the fuel (wood) could not be scientifically proved to be below the 400 degrees Fahrenheit threshold for the ignition of most types of wood and the tests of the ballasts had shown that even with a failed thermal protector, a ballast could not generate temperatures anywhere near that range. Their hypothesis would have to be based upon either an unreliable theory (pyrophoric carbon) or unsubstantiated assumptions and speculation about the temperature of the ballast which contradicted their own test results. As such, their testimony could not be admitted.

The appellate court affirmed the ruling of the trial court that the testimony of all of the experts had been shown to be not sufficiently reliable to be admitted under the standards of the Daubert decision. Without the testimony of the experts, Truck Insurance Exchange could not make a prima facie case for establishing the cause of the fire. Accordingly, the trial court entered Summary Judgment in favor of MagneTek and the appellate court affirmed this decision, as well.

This decision has significant implications for the litigation of fire cases everywhere. It demonstrates the importance of developing sound and scientifically verifiable theories for proving the cause of a fire as required by the standards of the Daubert decision. Moreover, it provides a compelling example of how a theory which

December 6, 2004 has not been validated and generally recognized by others in the scientific community will not withstand a Daubert challenge in court.

The lessons from this decision are many. First and foremost, experts must be prepared to prove the reliability of their investigative methodologies and theories to the satisfaction of the trial court. Experience alone is not sufficient. Even a theory which appears on the surface to be a reasonable and logical theory for the cause of a fire must be shown to be scientifically verifiable. Without a foundation in science, even the most experienced investigator will never be allowed to testify in court. For the parties hiring those investigators in their cases, there must be an awareness of the requirements for proving a case under investigative methodologies and theories which will meet the reliability standards of the Daubert decision, to guide them both in the selection of the expert used to investigate the fire and in the decision to litigate the case. For the attorneys handling those cases, there must be an awareness of those issues in order to successfully litigate that case at trial. The investigator, the client and the attorney each have a responsibility to ensure that cases are properly investigated and properly litigated. Without that awareness and recognition, the MagneTek case is a striking example of the consequences to be faced.

Citation: Babrauskas, V., Truck Insurance v. MagneTek: Lessons to Be Learned Concerning Presentation of Scientific Information, Fire & Arson Investigator 55:2, 9-10 (Oct. 2004).

Truck Insurance v. MagneTek: Lessons to Be Learned Concerning Presentation of Scientific Information by Vytenis Babrauskas, Ph.D.

Recently, a U.S. Court of Appeals decision was made in Colorado1 which is very unfortunate and which may make it more difficult for fire investigators to deal with ignition of fires which have been caused by long-term, low- temperature heating of wood. I have not had personal involvement with this case and know of the facts only as presented in the Court decision. Thus, my objective here is not to discuss whether justice was served for the parties involved, but, rather, to point out some implications for fire investigators and to offer some ways of coping with the new legal impediments that have been created.

The case concerned a ceiling which may have been set on fire by a fluorescent light ballast. The Court did not understand the difference between short-term heating of wood, where the concept of a roughly-known ignition temperature is applicable, and long-term, low-temperature heating behavior. When cellulosic substances which are capable of smoldering are exposed to sustained heating at temperatures which are much lower than the ‘ignition temperature’ applicable to short-term heating, under some conditions they may still ignite. This phenomenon is not newly-discovered or scientifically disputed. A scientific treatise2 was already published in 1984, although it is so highly mathematical that it would not be comprehensible to most investigators. Instead, the Court took as fact that wood has an ignition temperature of “approximately 400ºF” [204ºC] and rejected the plaintiffs’ case out of hand, because they could only demonstrate that, under normal conditions, the ballast would reach 232ºF [111ºC] and, under failure conditions, 340ºF [171ºC]. But the fact that smolderable combustibles may become ignited at temperatures much lower than an ignition temperature obtained during short-term heating has been known for over 100 years, as demonstrated by 19th century studies on spontaneous-combustion fires onboard ships carrying coal3.

The Court specifically rejected the plaintiffs’ claims because it concluded that the plaintiffs were presenting what it called a “pyrolysis theory.” It then went on to state that it found “the long-term, low-temperature ignition theory” to be “unreliable” and, therefore, excludable under Daubert. The implication to fire investigators now is that any time they encounter a case involving the long-term, low-temperature ignition of wood, their presentation might be thrown out of court. To avoid this, it is important to realize how the scientific presentation was made in the MagneTek case (or at least how it was interpreted by the court), and to make presentations in future cases which avoid such an out- of-hand rejection by the court.

“Pyrolysis” is not a theory, it is a definition. As such, it cannot be right or wrong (but, as with any definition, can be adhered to widely, or not). Analytical chemists define4 “pyrolysis” as “Transformation of a compound into one or more other substances by heat alone, i.e., without oxidation”. If oxygen does play a role, then chemists refer to the process as “oxidative pyrolysis,” and the latter term is widely used in the thermal-analysis branch of chemistry. In fire science, we normally define “pyrolysis” as stated in my Ignition Handbook5: “The chemical degradation of a substance by the action of heat.” Thus, the slight difference is that, in fire science, we include both oxidative and non-oxidative pyrolysis in the general definition.

But “pyrolysis” does not explain anything concerning long-term, low-temperature ignitions of wood. It is an observable fact that wood is a material which will pyrolyze, as opposed to, say, gold, which can be raised to quite high temperatures without suffering chemical degradation. That fact, by itself, says nothing as to whether it can even ignite, much less under what conditions it will ignite. To determine under what conditions something can ignite, in principle, we can do one of two things: (1) use a scientific theory to make a calculation; or (2) consult observational data collected on the topic.

1 Truck Insurance Exchange v. MagneTek, Inc. (No. 03-1026), U.S. Court of Appeals, Tenth Circuit, Appeal from the U.S. District Court for the District of Colorado, D.C. No. 00-RB-2218(CBS), Filed Feb. 25, 2004. 2 Bowes, P. C., Self-Heating: Evaluating and Controlling the Hazards, Her Majesty’s Stationery Office, London (1984). 3 Rowan, T., Coal Spontaneous Combustion and Explosions—Occurring in Coal Cargoes—Their Treatment and Prevention, E&FN Spon, London & New York (1882). 4 Lewis, R J. sr., Hawley’s Condensed Chemical Dictionary, 14th ed., p. 941, Wiley, New York (2001). 5 Babrauskas, V., Ignition Handbook, Fire Science Publishers/Society of Fire Protection Engineers, Issaquah WA (2003). 2

The problem with the plaintiffs’ case was that they evidently tried to provide a theory, which did not serve them well. There are many events in the natural sciences which are well documented, but lack a theory. We have very good documentation that Mount St. Helens last erupted on May 18, 1980. We have no theory that can explain why it erupted on May 18th, instead of May 15th. But that fact does not stop us from being entirely, scientifically correct when we point out that Mount St. Helens did erupt on May 18, 1980. The same tack should have been taken in the present case. As documented extensively in the Ignition Handbook (which, by the way, was peer-reviewed and was published under the aegis of the Society of Fire Protection Engineers), ignitions due to long-term, low-temperature heating of wood members are documented down to a temperature of 170ºF [77ºC]. For temperatures above 212ºF [100ºC], documentation is not only reliable, but copious. Based on these facts, it should have sufficed for the plaintiffs to demonstrate that the product imposed temperatures on wood in excess of these. If this is demonstrated, then it directly follows that such a product is unsuitable for use where it would be in contact with wood for an extended period of time, and that it creates an imminent danger of fire. This, of course, must be accompanied by good fire investigation which excludes other potential energy sources as an ignition source for the fire.

The state-of-the-art of fire science is that we do not have a model that can predict this phenomenon. That is to say, if the scientist is told that the piece of wood has a certain size and configuration, and the hot object has certain known characteristics, he cannot make computations along the lines of “It will take 5.9 months for fire to break out.” The physics and chemistry (and biology!) of wood are very complicated. We have a certain understanding of factors that play a role in this, but the understanding is nowhere near progressed to where numerical computations can be made. The long time scale involved and the fact that cracking evidently plays a crucial role (but is extremely hard to model) are factors that contribute to the difficulty. Computations on the spontaneous combustion behavior of some other substances, where the phenomena are less complicated (for example, powdered milk6) have been made. The size/temperature relations needed for powdered milk to ignite can reasonably be estimated, although even in the best of circumstances estimating the time interval—which may be of greatest importance to the fire investigator—is hard to do reliably, because existing theories differ widely. The mathematical details of this are discussed in the Ignition Handbook.

Hopefully, the next time the issue arises in court, the expert who is presenting the material will be armed with the knowledge that, in the general case, scientific truth can be demonstrated either by using a theory or by pointing to the collected set of observations on the topic. But in the case of long-term, low-temperature ignition of wood, the first option is not available, so the second one perforce must be used. This does not devalue the presentation. Despite the incorrect view of some laymen that “science” means “a collection of equations,” this is not true. The basic dictionary definition7 of science is: “A branch of study that deals with a connected body of demonstrated truths.” Geology is a good example of a science where there are many facts that have been demonstrated to be true, but few equations exist for calculating anything.

Finally, since a theory of long-term, low-temperature ignition of wood is not available, obviously it should be urged that research be carried on so that one day this might become possible. This will have to be preceded by laboratory experiments. In that connection, a caution must be offered. Because of the microscopically-irregular nature of wood, and the fact that cracking plays a role, one has to be prepared for a wide range of data spread. In other words, if a number of experiments were to be made where a wood member is ignited from a hot surface applied for a long time, the expectation is that the times to ignition will show a wide range of values. In addition, it will be found that some specimens will fail to ignite, while others, held at conditions as identical as possible, ignite. This has been observed for other substances where extensive laboratory tests were made.

6 Beever, P. F., Spontaneous Ignition of Milk Powders in a Spray-Drying Plant, J. Society of Dairy Technology 37, 68-71 (1984). 7 The New Shorter Oxford English Dictionary, Clarendon Press, Oxford (1993). 3

(SIDEBAR) NFPA 921 guidance on the topic

NFPA 921 points out (Sec. 5.3.1): “If the fuel is to reach its ignition temperature, the heat source itself must have a temperature higher than the fuel’s ignition temperature. Spontaneous ignition is an exception.” The correct message is conveyed although the chosen term, “spontaneous ignition,” is not the best, because in a later section (Sec. 5.3.2.5) the same term is used in an entirely different sense, meaning unpiloted ignition during short-term, external heating. To make matters worse, Sec. 5.3.6.1 describes spontaneous combustion due to self-heating and states that “In this type of reaction, self-heating of a material to its ignition temperature will result in self-ignition.” This is misleading because, if a material ignites as a consequence of long-term, low-temperature heating, the concept of “its ignition temperature” is wholly incorrect, since the temperatures involved are very sensitive to size, geometry, and ventilation conditions and there can be no handbook temperature which is a property of the fuel alone. But the 2004 edition has helpfully added a new section (Sec. 5.3.6.2.7) specifically on wood, stating that: “Exposure temperatures needed for wood self-heating to ignition are significantly lower than those shown in Table 5.3.5 for flaming ignition of fresh wood.” Unfortunately, another new section, Sec. 18.3.2 states: “A competent ignition source will have sufficient temperature and energy and will be in contact with the fuel long enough to raise it to its ignition temperature,” overlooking self-heating as a competent source of ignition.

Dr. Vytenis (Vyto) Babrauskas has degrees in Physics and Structural Engineering and was the first- ever person to be awarded a Ph.D. degree in Fire Protection Engineering (Univ. California, Berkeley, 1976). For 16 years he was at NIST, where he conducted laboratory studies on fire hazards and developed engineering design methods. In 1993 Dr. Babrauskas founded Fire Science and Technology Inc., specializing in fire science support for fire investigations. He is noted for having invented the Cone Calorimeter and the furniture calorimeter. Dr. Babrauskas has published over 250 papers and reports and three books: Heat Release in Fires, Fire Behavior of Upholstered Furniture and Mattresses, and the new Ignition Handbook. The latter is the first handbook on the subject of ignition and covers both scientific aspects and practical advice for fire investigators.

In 2001, prior to the completion of the Ignition Handbook, the following paper was published, giving some interim findings: Babrauskas, V., Pyrophoric Carbon…The Jury is Still Out, Fire and Arson Investigator 51:2, 12-14 (Jan. 2001). Due to the interim nature of this initial publication, below is a revised presentation of this topic which takes into account the final findings, as published in the Ignition Handbook. The Handbook is a peer-reviewed publication and was published under the auspices of the Society of Fire Protection Engineers in 2003.

‘Pyrophoric Carbon’ and Long-term, Low-temperature Ignition of Wood

by Vytenis Babrauskas, Ph.D.

Summary Recently, claims have been made that scientific research has disproven the concepts of ‘pyrophoric carbon’ or ‘pyrophoric char.’ Because of this, it has been claimed that wood exposed to long-term, low-temperature heat sources cannot exhibit spontaneous combustion, if the heat sources are below about 150ºC (about 300ºF). It will be shown that these claims only arise because an oversimplified theory was used, which is not capable of quantifying important phenomena in the problem. It will also be shown that the experimental research in support of those conclusions did not, in fact, speak to the issue. Finally, it will be shown that (a) practical guidance is available concerning conditions that are hazardous, with respect to long-term, low-temperature ignition of wood, but that (b) additional research is desirable in order to obtain an understanding of the physicochemical details of this phenomenon. The available practical guidance—i.e., the fires that have been documented to have occurred when wood members were exposed to heating sources at 77ºC (170ºF) or higher—forms a reliable, scientific basis for concluding that an ignition hazard exists if a heat source at 77ºC or higher is applied to a wood member for a protracted period of time.

Background—The scientific concepts of ignition Ignitions of combustible substances are generally divided into two types: 1. due to external heating 2. due to internal heating (self-heating). There are situations of overlap, but this division proves convenient both in the scientific world and in practice. Most ignitions encountered are due to external heating. If a match ignites paper, or a halogen lamp ignites a curtain, the source is external. Fire investigators are also well familiar with internal heating. When a pile of linseed-oil soaked cloths bursts into flame, the cause is due to internal heating. Internal heating is more complicated than external heating, since factors such as air flow or exact details of the oxidation chemistry, which are of minor importance in external ignitions, become of major concern.

There had earlier been much disagreement about the minimum temperature needed for ignition of wood due to external heating. A recent study1, however, systematically examined the results from more than a century’s worth of research on this topic and concluded that—if wood is heated under the minimum heating conditions that will suffice for it to ignite—it will ignite at a temperature of approximately 250ºC (482ºF). If wood is exposed to an environment where heating conditions are greater than the minimum ones that are required, the actual temperature of ignition rises; thus the 250ºC value is a suitable limit for design or hazard analysis purposes.

If a wood member ignited upon being heated for minutes, or up to a few hours, then it can be concluded that it ignited due to external heating. However, if wood is exposed for months or years to a temperature lower than 250ºC, it can still ignite, because, on this time scale, self-heating plays a role. The recently-published Ignition Handbook2 contains a collection of data on the temperatures that were involved in various real-life incidents where wood members ignited after exposure for months-to-years-long time periods. A particularly well-documented incident described in the Handbook of this phenomenon is shown in Figure 1. The Handbook also describes that the incident involving the lowest documented temperature at which a fire was reported involved a hot-water pipe operating at 77ºC (170ºF). 2

(a) Ignition of wood floor/ceiling assembly from a hot-water supply pipe (left pipe in the photo) (Courtesy Ken Swan)

(b) Close-up of above (Courtesy Ken Swan) Figure 1. Fire caused by contact of a supply pipe from a hot-water heating system. The pipe penetrated a floor assembly comprising a 38.1 mm (1.5") wall plate on top of a 15.9 mm (5/8") OSB subfloor; the furnace was producing water at a temperature of 88 – 93ºC.

It is important to note that since the physical and chemical mechanisms involved in ignitions due to external heating, and those due to self-heating are different, no specific relation can be assumed to exist between the 250ºC value and the 77ºC value.

A theory for the self-heating of substances was first developed by the Russian researchers Nikolai Semenov and David Frank-Kamenetskii from 1928 through the 1940s. The theory is generally called the “F-K theory” in honor of Frank-Kamenetskii. The theory was subsequently refined by British researchers, especially Philip Thomas, and came into fairly wide use by the 1960s. To understand some of the uses of this theory, it is first important to understand some basic terms. Self-heating: an increase in temperature due to exothermicity of internal reactions. Thermal runaway: self-heating which rapidly accelerates to high temperatures. Spontaneous combustion: visible smoldering or flaming caused by thermal runaway. 3

Almost all organic substances (and many inorganic ones) can undergo exothermic reactions, typically, decomposition or oxidation. Since that is the case, it is necessary to understand why these substances are not spontaneously igniting all the time. The answer lies in the difference between self-heating and spontaneous combustion.

Suppose a modest heap is piled up of milk powder, a substance known to exhibit self-heating tendencies. If the pile is not too large, the temperature inside will rise by maybe 10°C, then will slowly drop back down. Suppose now that a very big pile of the milk powder is heaped together. The temperature inside will rise slowly at first, then start to accelerate very rapidly. The material will be the hottest in the inside. It will start to smolder rapidly and the smolder front will advance through the material. Finally, flaming may break out when the smolder front reaches the outside surface. In other cases, the entire pile may be consumed by smoldering, and flaming will not appear.

It has been found that the F-K theory can reasonably represent the events that occur in the pile of milk powder3. The theory, however, is based on extremely simplified chemistry and physics: • There is only a single chemical reaction that takes place, and this reaction can be represented by an equation of the so-called Arrhenius form. • All reaction products are the final products, and no further reactions amongst the products need to be considered. • As much oxygen is available as needed. • No account is taken of the mass flow of anything. • Moisture has no role in the process. • No physical changes, for example, cracking, are considered. With these simplifications, it may be surprising that the theory can predict anything! But, in fact, there have been some well-publicized successes. On the other hand, it has been found out that the F-K theory does not apply to: • haystacks (due to biological heating by micro-organisms) • bagasse piles (due to pivotal role of moisture flow) • coal (due to complex interrelation between oxygen and moisture) • fertilizers (due to multiple reactions, high variations in thermal conductivity, moisture flow, and melting) • numerous other substances2, often of ones industrial concern.

The historical problem of low-temperature, long-term ignitions of wood Around 1900, fires started being reported with steam or hot-water heating pipes where the pipes had been passed through wood members. Ignitions were being observed typically 3 months to 15 years after installation4,5. The installations typically involved hot-water or low-pressure steam, where temperatures should be not much over 100°C (212°F). A small fraction of the incidents evidently involved some boiler malfunctions, and notably high temperatures would be expected. But for the majority, it is clear that these ignitions are not be categorized as external ignitions. It might be noted that investigators making the determinations of the cause in these long-term, low-temperature wood ignitions included Prof. Ira Woolson6 (the first US professor of fire science) and Voitto Virtala7 (the ‘father’ of fire science in Finland).

Even in the era of 100 years ago, it was evident that such ignitions were, in some sense, different from external- heating ignition and that a different explanation had to be provided. The original explanation that was proposed for these ignitions was put forth by the German scientist Ernst von Schwartz8 in 1902. Based upon a theory earlier suggested by the German scientist H. von Ranke as an explanation for haystack fires, the theory claims that low- temperature, long-term heating of wood converts the wood to “pyrophoric carbon,” and that this pyrophoric carbon is much more readily ignitable than is virgin wood.

For most of the 20th century, this theory was taken for granted, but no research was done to quantify or to understand it. Dr. Fred Shafizadeh, professor at University of Montana, and considered to be the world’s leading authority on pyrolysis of wood, started investigating the problem in the late 1970s. Shafizadeh created wood chars under aerobic (having access to air) and anaerobic (inert atmosphere) conditions. He found that char created under anaerobic conditions was highly reactive. The reactivity was an exothermic chemisorption of oxygen onto the char surface. Shafizadeh stated9: “This gives credibility to Ranke’s theory of pyrophoric char and his explanation of spontaneous ignition.” Unfortunately, Prof. Shafizadeh died while the research was still in progress.

4

It has been often reported that 288ºC (550ºF) is the charring temperature of wood. But this has been based solely on very brief exposures. In fact, it does not require high or even medium temperatures for wood to get charred—this has been well documented although less publicized. In 1945 McNaughton10 reported on a study done at the US Forest Service laboratories, where he exposed tiny (3 mm × 6 mm) sticks of wood for long periods. His results are shown in Table 1. Table 1 Results of exposure of matchstick-size wood samples for long periods Temperature Time Condition (days) 107°C 1050 light chocolate color 120°C 1235 brittle, dark-chocolate color 140°C 320 lost 45% of weight, became like charcoal 150°C 165 lost 65% of weight, became like charcoal

These results prove that very small pieces of wood, when exposed for a few years to 120 – 150°C temperatures, will turn into char. A number of fires have occurred, however, at temperatures below 120ºC, and as low as 77ºC. It is also obvious that wood will not ignite due to self-heating unless it has actually charred. Thus, it is important to understand that the McNaughton study only documents the starting-point temperature needed to cause charring for extremely small pieces of wood, much smaller than ones that have been involved in the known fires. The reason that temperatures as low as 77°C were sufficient to char and to ignite wood members is because the members involved were substantially larger than McNaughton’s matchstick-sized specimens. Even though there does not currently exist a computational formula which would allow direct prediction of the ignition of such wood members based on a size/temperature relationship, it is well-known in the scientific literature that, for all the substances that have been studied, a lower starting-point temperature suffices if the size of the specimen is increased. Thus, it is not surprising that members which are a 2" × 4" (nominal) size, or larger, have been found to char and to ignite at lower starting- point temperatures than matchstick sized test specimens.

Efforts to disprove long-term, low-temperature ignitions of wood There have been two efforts in recent times to prove that long-term, low-temperature ignitions of wood cannot occur. In 1984, Philip Bowes, a researcher at the Fire Research Station in the UK, published his conclusions11 that wood members cannot be ignited from steam pipes that are operating at less than 200°C. Bowes did not do any experimental work to address the specific question. Instead, he analyzed existing experimental data in an attempt to find the answer. His basic reasoning steps were the following: • He showed that wood samples which were pre-charred in an atmosphere where the oxygen availability was never limited, and then exposed to external radiant heating (in unlimited oxygen), ignited at about the same temperature as did non-pre-charred wood. • Using available data obtained from tests of small cubes of material in an oven with unlimited access to oxygen, he applied the F-K self-heating theory. • The theory indicated that a steam pipe would have to be operating at a temperature of about 200°C for runaway conditions to take place. • Thus, he concluded that all of the reported investigations of fires which were attributed to sub-200°C heat sources had been incompetent, and that the steam or hot-water pipes actually were at temperatures of at least 200°C. As described in detail in the Ignition Handbook2, at the present time there is an incontrovertible collection of data to indicate that competent fire investigators have documented fires due to this cause at temperatures much lower than 200ºC, and in fact as low as 77ºC.

More recently, in 1999, Bernard Cuzzillo published the findings from his dissertation12,13. In his thesis, he claims that he has disproven the existence of pyrophoric carbon (but it is not entirely clear from his writings whether he holds the same position as Bowes, or whether he accepts that long-term, low-temperature ignitions of wood do occur, but that they occur due to some mechanism which does not include oxidation of a reactive char). Cuzzillo based his thesis on three series of experiments that he conducted: (1) tests in an oven, with unlimited oxygen, on chips of virgin wood (2) tests in an oven, with unlimited oxygen, on chips which had been pre-charred under unlimited oxygen conditions 5

(3) tests in an oven, with unlimited oxygen, on 89 mm cubes of whole wood. His tests indicated that: • there was no significant effect of pre-charring on the behavior of the wood chips • a wood member of 89 mm size must be exposed to a temperature of around 200°C to go into thermal runaway. Cuzzillo then used the F-K theory as a tool to extrapolate the data to other sizes. But since an 89 mm (3.5") size is fairly typical for beams through which a steam pipe might imprudently be cut through, there was not a great deal of need for extrapolation.

Analysis Both Bowes and Cuzzillo succumbed to a similar flaw in logic: Both relied on data collected in test environments dissimilar to the environment of actual fires. Both used extrapolations of, or numerical calculations with, a highly simplistic theory. Calculations with the F-K theory should not even have been necessary, since simple plots of existing data from oven-tested specimens would have shown that the thermal-runaway time scale for, say 90 or 150 mm size specimens is hours, not months or years. This should have suggested to Bowes and Cuzzillo that the F-K theory cannot hope to explain ignitions of modest-sized members that take months or years to occur.

It is important to consider Shafizadeh’s view that highly-exothermically-reactive (that is, ‘pyrophoric,’ although the term is not a well-chosen one, in the opinion of this author) char is formed under conditions where oxygen is excluded from the char generation site. In his textbook, John DeHaan14 points out that fire cases identified as being due to pyrophoric carbon are especially prone to occur when the wood is sealed under an impervious layer, such as sheet metal. The generation of reactive char occurs initially under conditions where access of oxygen is limited. Subsequently, something happens that makes oxygen more available to the reactive char. DeHaan illustrates this with cases where a metal or tile covering is originally present, then becomes disrupted. It can also be envisioned that the process may well occur in the case of the ordinary steam pipe. Steam pipes and heating ducts play a similar role to a sheet-metal cover, in that they occlude the flow of oxygen for quite a while, until substantial shrinkage or charring of wood occurs. It is known that wood char cracks as heating continues. The time scale of the problem is most likely established not by a simplistic F-K one-step chemistry reaction, but by the char-cracking process. This is, unfortunately, a very poorly understood process, but it may well be playing a crucial role.

Thus, it is entirely likely that Prof. Shafizadeh was right in hypothesizing that long-term ignitions are a 2-step process: (1) a reactive char gets formed under restricted-oxygen conditions. (2) the reactive char then ignites. This may occur when further shrinkage takes place and oxygen enters newly- formed cracks. But a number of things are not known about the process: • The flow of oxygen in depth into a wood member exposed to heat, but having only limited access to air. • The mechanics of char, that is, being able to predict when and how it cracks. • The relation between chemical reactivity and formation of cracks. • The role of moisture in the entire process. Prof. Woolson and a number of later investigators have suggested that cyclic heating appears to be more deleterious than a constant temperature. The reason for this is not currently known, but clearly moisture will be one of the main variables that undergo large changes when thermal cycling occurs. It may be surprising, but the details how wood char cracks have received only very brief study15,16 and the studies do not provide a basis for answering the above questions.

Conclusions For wood that is exposed to heat sources for no longer than a few hours, the concept of external ignition is applicable. Self-heating plays a negligible role at short durations, and an evaluation of the results from numerous experimental studies of ignition under external-heating conditions shows a mean value of 250ºC (482ºF) for the ignition temperature obtained under worst-case (minimum heat flux) conditions. For wood exposed to lower temperatures, but for long time periods (months-to-years), self-heating is the dominant phenomenon. The concept of a fixed (“handbook”) ignition temperature does not apply to a self-heating substance, and the hot-object temperature which will suffice to cause ignition under such circumstances depends strongly on the size of the specimen. For wood building members exposed to such long-term heating conditions, case incidents indicate that ignition is possible at a hot-object temperature as low as 77ºC (170ºF). It must be emphasized that the hot-object 6 temperature is the starting-point temperature of the self-heating process and that the self-heating process progressively raises the temperature of the wood member, so that when it actually ignites it will be at a temperature that is much higher than the starting-point temperature.

In terms of safe design and safe practices for the installation of heat-producing devices adjacent to wood surfaces, it should not be a new or surprising piece of information that 250ºC would represent an extremely hazardous condition and that 77ºC, in fact, must not be exceeded if the heating is prolonged. Already in 1959 UL17 issued this recommendation: “As a limitation on the temperature to which wood may be heated for long periods of time from a standpoint of fire prevention, many authorities indicate that 90ºF above room temperature (approximately 80ºF) normally prevailing in habitable spaces is a safe maximum and one which incorporates a reasonable margin of safety.” Since 80 + 90 = 170ºF, the temperature cited in the 1959 UL recommendation is identical to the one derived in the present study. But there is an important difference in that, in our study, the 77ºC value has a zero safety factor, whereas the UL value of 77ºC, based on much earlier research, was intended to include some positive, but unspecified, safety factor. In any efforts to establish an “allowable” temperature, when considering data on temperatures at which ignition can occur, a safety factor must be included to ascertain that ignition will not occur, rather than will-just-barely occur.

No researcher, on the basis of applying a theory (especially a theory which is too simplified to capture the actual chemistry and physics of the phenomenon) can disprove the possibility of long-term, low-temperature ignitions of wood, since credible documentation, reported by reliable sources, exists to characterize fires due to this cause. A theory which disagrees with factual observations is obviously to be rejected. The studies of Bowes and Cuzzillo did not improve the theory beyond the already-available F-K theory, which is entirely inadequate to encompass the chemistry and physics involved in the present case. In addition to theory, Cuzzillo also conducted laboratory testing, but since the tests were performed under conditions notably different from those prevailing in real-life, long-term, low-temperature ignitions of wood, these results do not help to quantify the conditions under which such ignitions will occur.

The chemistry, physics, and thermostructural behavior involved in producing spontaneous combustion due to long- term, low-temperature heating of wood are clearly very complicated, interrelated phenomena. It may be many decades before a theory is evolved that can give useful numerical results. However, expanded guidance to investigators could be obtained in a few years by conducting a series of well-designed experiments. The simplest and best strategy might involve the following: • Conduct real-scale tests (e.g., actual wood members) under realistic test conditions. This means a mockup of a meter or so in size, with maybe a 150 mm (6") thick member being used. These should not be oven tests, but, rather, the wood member must be heated the way it will be in practice—by a hot metal pipe or duct. • In view of the possible importance of thermal cycling, two types of tests should be run: ones with the pipe at a constant temperature, and ones with temperature cycling. • A period of 3 years might be a suitable time during which the experiment should be in place and left to run its course.

The US Dept. of Transportation18 defines: “A pyrophoric material is a liquid or solid that, even in small quantities and without an external ignition source, can ignite within 5 minutes after coming in contact with air.” Thus, applying the term ‘pyrophoric’ as a descriptor of long-term, low-temperature ignition of wood is not appropriate.

Finally, in view of the recent confusion in the Courts19, it is essential to emphasize that scientific knowledge of a phenomenon does not require that a theory or an equation exist for it. It is obviously the desire of all scientists to increase knowledge to the point that reliable computational theories become available. But the fundamental basis of all science is a systematic collection of data. And a collection of observations is sufficient to form a basis for making scientific conclusions. Thus, in the present case, it is possible to reliably conclude that any heating device of 77ºC or higher, if applied to a wood surface for a protracted period of time, presents a documented ignition hazard. 7

References

1. Babrauskas, V., Ignition of Wood: A Review of the State of the Art, J. Fire Protection Engineering 12, 163-189 (2002). 2. Babrauskas, V., Ignition Handbook, Fire Science Publishers, and Society of Fire Protection Engineers, Issaquah WA (2003). 3. Chong, L. V., Shaw, I., and Chen, X. D., Exothermic Reactivities of Skim and Whole Milk Powders as Measured using a Novel Procedure, J. Food Eng. Vol. 30, 185-196 (1996). 4. Bixel, Edward C., and Moore, Howard J., Are Fires Caused by Steam Pipes? (B. S. thesis), Case School of Applied Science, Pittsburgh (1910). 5. Matson, A. F., Dufour, R. E., and Breen, J. F., Survey of Available Information on Ignition of Wood Exposed to Moderately Elevated Temperatures, Part II of Performance of Type B Gas Vents for Gas-Fired Appliances (Bull. of Research 51), Underwriters’ Laboratories, Inc., Chicago (1959). 6. Low Temperature Ignition of Wood, NFPA Q. Vol. 19, 159-167 (1925). 7. Virtala, V., Oksanen, S., and Fridlund, F., Om självantändlighet, dess bestämning och förekomst [On spontaneous ignition and its occurrence; methods for the determination of the tendency to spontaneous ignition], (Julkaisu 14), Valtion Teknillinen Tutkimuslaitos, Helsinki (1949). 8. Schwartz, E. von, Fire and Explosion Risks, translation of original German text. Charles Griffin & Co., Ltd., London (1904). 9. Bradbury, A. G. W., and Shafizadeh, F., Role of Oxygen Chemisorption in Low-Temperature Ignition of Cellulose, Combustion and Flame Vol. 37, 85-89 (1980). 10. McNaughton, G. C., Ignition and Charring Temperatures of Wood, Wood Products Vol. 50, 21-22 (1945). 11. Bowes, P. C., Self-Heating: Evaluating and Controlling the Hazards, Her Majesty’s Stationery Office, London (1984). 12. Cuzzillo, B. R., and Pagni, P. J., Low-Temperature Wood Ignition, Fire Findings Vol. 7, No. 2, 7-10 (1999). 13. Cuzzillo, B. R., Pyrophoria (Ph.D. dissertation). University of California, Berkeley (1997). 14. DeHaan, J. D., Kirk’s Fire Investigation, 4th ed., Brady/Prentice-Hall, Englewood Cliffs, NJ (1997). 15. Phillips, S. A., How Wood Chars and What It Means to the Fire Investigator, Fire and Arson Investigator Vol. 38, No. 4, 28-30 (June 1988). 16. Ettling, B. V., The Significance of Alligatoring of Wood Char, Fire and Arson Investigator Vol. 41, No. 2, 12- 15 (Dec. 1990). 17. Matson, A. F., Dufour, R. E., and Breen, J. F., Survey of Available Information on Ignition of Wood Exposed to Moderately Elevated Temperatures, Part II of “Performance of Type B Gas Vents for Gas-Fired Appliances” (Bull. of Research 51), Underwriters’ Laboratories, Inc., Chicago (1959). 18. Class 4, Divisions 4.1, 4.2, and 4.3—Definitions, Code of Federal Regulations, 49 CFR 173.124 (1998). 19. Babrauskas, V., Truck Insurance v. MagneTek: Lessons to Be Learned Concerning Presentation of Scientific Information, Fire & Arson Investigator 55:2, 9-10 (Oct. 2004). Babrauskas, V., Ignition of Wood: A Review of the State of the Art, pp. 71-88 in Interflam 2001, Interscience Communications Ltd., London (2001).

IGNITION OF WOOD A REVIEW OF THE STATE OF THE ART

Vytenis Babrauskas, Ph.D. Fire Science and Technology Inc., 9000 – 300th Place SE, Issaquah WA 98027, USA

This review encompasses the available practical and experimental data on the ignition of solid wood. Only solid, natural wood is considered, not sawdust, chips, or products that have been treated with fire retardants or other substances, nor is the ignition of living trees. Panel products such as plywood or particleboard have ignition properties very similar to solid wood, so the solid-wood results will generally be applicable to them. Wood may ignite by flaming directly, or it may ignite in a glowing mode, which may or may not be followed by flaming. It is shown that the ignition temperature is around 250ºC for wood exposed to the minimum heat flux possible for ignition, and that it invariably ignites, at least initially, in a glowing mode under these conditions. The ignition temperature rises rapidly as the heat flux is increased. Piloted ignition at heat fluxes sufficient to cause a direct-flaming ignition normally occurs at surface temperatures of 300 – 365ºC. Autoignition temperatures at fluxes higher than minimum are essentially unknown. No theory is available which encompasses the possibility of glowing, glowing followed by flaming, or direct-flaming ignition modes. Most published studies have dealt with radiant or radiant+convective heating, and knowledge is extremely poor for ignition from direct contact by hot bodies or by flames. A species- independent correlation is derived for the radiant, piloted ignition of thermally-thick wood, but the fit is only fair. The minimum flux for ignition is 4.3 kW m-2, based on a single study; most reported tests have been much too brief to produce useful data on this point.

IGNITION TEMPERATURE The concept that combustible substances ignite when a given surface temperature is first attained is an empirical notion—in many cases, this is found to be true enough, so that even though not exactly true, the concept has utility and merit. It has also found significant application to theoretical modeling—closed-form theories for radiant ignition, for example, generally assume that ignition corresponds to a known, constant surface temperature Tig. Thus, the starting point for investigating the ignition of wood must be to examine experimental data on its ignition temperature. As can be seen in Table 1, studies on this question go back well into the 19th century and have continued until the present time. The spread of data is clearly enormous. It might first be noted that even the term ‘ignition temperature’ tends to mean two different things: (1) the temperature of the surface at the time of ignition; or (2) the minimum temperature of a furnace sufficient for a specimen put therein to ignite. The latter notion might seem to be old and non-rigorous, but it must be remembered that: (a) the common test for ignition temperature is the Setchkin furnace, ASTM D 1929 [1], which is based on the latter definition; and (b) the user often needs to know the highest environment temperature to which he can subject a material without it igniting and he may be less interested in actual temperatures at the specimen. Excluding one value, the results in

71 Table 1 span 210–497ºC for piloted ignition and 200–510ºC for autoignition. The following reasons should be considered that might account for the spread: • the definition of ignition that is used • piloted vs. autoignition conditions • the design of the test apparatus and its operating conditions • specimen conditions (e.g., size, moisture, orientation) • species of wood. The definition of ignition is complicated not only by the two meanings currently in use, but by some practices followed by earlier investigators. Until the 1960s or so, it was not rare for investigators to report ignition results without making visual observations. Strange as this may seem from today’s perspective, a number of studies exist where the ignition criterion was based solely on thermocouple readings. Typically, the test rig was equipped with two thermocouples and a criterion was used which related the value or slope of the one reading to the other. Results of this kind might be automatically excluded from consideration, except for the fact that data from those investigators do not seem to be systematically different from the others. Table 1 Summary of ignition temperature results for wood Year Investigator Spec. Ignition Comments size temperature (ºC) Piloted Auto- ignition 1887 Hill [2] 0.5-15 g 220–300 measured air temperature near sample 1910 Bixel, Moore [3] 35 mm ? 200–250 measured oven temperature; scant details 1922 Banfield, Peck [4] 50×50× 302–308 measured surface temperature 200 mm 1934 Brown [5] 1–5 g 220–250 measured oven temperature; tiny samples; unsound ignition criterion 1936 VanKleeck [6] chips 235 measured specimen temperature; unsound ignition criterion 1947 NIST [7] shavings 228–264 softwood shavings in test tube; criterion—glowing or flaming 1949 Graf [8] 7–13 g 232–245 measured oven temperature; tiny samples; unclear ignition criterion 1949 Angell [9] 13‰19‰ 204 measured gas temperature close to specimen 51 mm 1950 Fons [10] 2-9 mm 343 measured oven temperature; solved inverse cylinders problem 1958 Narayanamurti ? 228 measured oven temperature [11] 1959 Thomas et. al. 32‰32‰ 210 measured oven temperature; solved inverse (data of Prince, 102 mm problem 1915) [12] 1959 Akita [13] 20×20× 450 489 measured oven temperature; solved inverse 1.8 mm problem < 350 measured oven temperature only 1960 Simms [14] 8 mm ø 525 calculated from correlation, not measured 1960 Moran [15] 50×50× 255 at flux = 25 kW m-2; measured surface 6.4 mm temperature 1961 Patten [16] 3 g 260 260 measured oven temperature (Setchkin test) shavings 1961 Buschman [17] 57×57× 369 calculated from correlation; fluxes 14.3 to 8 mm 37.2 kW m-2 1964 Shoub, Bender 920×920 254 measured surface temperature

72 [18] mm 1964 Tinney [19] ≥6 mm ø 350 measured oven temperature 1967 Simms, Law [20] 76×76× 380 calculated from correlation 19 mm 1969 Melinek [21] 100×100 353 382 calculated from correlation ×13 mm 1969 Jach [22] few 260–290 measured oven temperature grams 1970 Smith [23] 75×75× 350 413–714 temperatures measured by optical 19 mm pyrometry; autoignition values dubious 1983 Atreya [24] 64 mm ø 370 temperatures measured, but below surface; × 19 mm flux = 18 kW m-2 350 temperatures measured, but below surface; flux ˆ 30 kW m-2 1986 Atreya et al. [25] 75×75 330–405 temp. measured, but below surface ×19 mm 1988 Abu-Zaid [26] 150×75 420 forced-air flow; temp. measured but below ×37 mm surface; flux = 18.5 kW m-2 350 forced-air flow; temp. measured but below surface; flux > 25 kW m-2 530 flux = 40 kW m-2 1991 Janssens [27] 100×100 300–364 surface temp. measured; ×17 mm fluxes 25 to 35 kW m-2 1992 Li, Drysdale [28] 64×64 411–497 temp. measured but below surface; ×18 mm flux < 20 kW m-2 353–397 temp. measured but below surface; flux > 20 kW m-2 1993 Masařík [29] 2.5 g 220–240 tested wood fiberboard; measured oven temperature (Setchkin test) 1996 Fangrat [30] 100×100 296–330 surface temp. measured; mm fluxes ˆ 25 kW m-2 1997 Moghtaderi [31] 100×100 332 temp. measured but below surface; at 20 ×19 mm kW m-2 297 temp. measured but below surface; at 60 kW m-2 ? – denotes unknown measurements

The design of the test apparatus has perhaps the largest influence. The majority of devices fall into one of two types: (1) a furnace into which a small specimen is bodily plunged; or (2) a specimen sitting in the open air and being radiatively heated, e.g., the Cone Calorimeter [32]. But this basic division is confounded by the fact that there is a preferred specimen type for each test: specimens of only a few grams are normally put into a furnace that exposes the whole specimen bodily, while specimens placed in front of radiant heaters are typically on the order of 100 g and of sizeable dimensions in at least two directions. The results are summarized in Table 2, with type 1 values indicated in bold in Table 1 and type 2 underlined.

Considering first autoignition temperatures under radiant heating, the results evidently span a huge range. Smith’s results (which go up to 714ºC) appear to be implausible and may refer to an average optically measured temperature on which some spots are already glowing; thus, they will be excluded. Several other workers reported calculated, rather than measured, values; these will be presumed to be less reliable. Of the measured values, Moran’s value of 255ºC and Shoub’s 254ºC are impressively close. The only other value obtained by actual measurement is Abu-Zaid’s 530ºC. But his result was obtained at a heat flux of 40 kW m-2, which is much higher than Shoub’s 4.3 kW m-2 or Moran’s 25 kW m-2. This suggests that different flux regimes must be considered. Thus, it might be assumed that 250ºC is

73 characteristic at very low fluxes, while some much higher temperature is obtained at high heat fluxes. Turning now to autoignition in ‘a few grams plunged into a furnace’ tests, if the range reported by each investigator is averaged, then the data span only 235–275ºC, with an average of exactly 250ºC. It may be noted that the ‘a few grams plunged into a furnace’ tests are normally operated in such a way as to only seek out the condition where the furnace temperature is the minimum for ignition. In principle, they can be run at non-minimum temperatures, but such data are hardly ever reported. Thus, from this type of test there is no corresponding result to the high-flux region of radiant tests. It can be concluded then that if a wood specimen is ignited under external heating barely sufficient to ignite it, it will ignite at ca. 250ºC regardless of the type of heating arrangement. Table 2 Summary of ignition temperature data Type of test Ignition temperature (ºC) Piloted Autoignition a few grams plunged into a furnace 220–260 220–300 radiant heating of a largish specimen 296–497 254–530 others; unidentified 210–450 200–525

Concerning autoignition at higher heat fluxes, the paucity of reliable data makes it difficult to draw useful conclusions. Simms’ calculated value of 525ºC is close to Abu-Zaid’s measured 530ºC, but both seem very high. Akita’s calculated value of 489ºC is lower, but his results appear to be too high (see below), so the actual value was probably lower yet. It also appears that apparatus details play a stronger role in autoignition than for piloted ignition, leading to wider scatter.

For piloted ignition, Tig values should not be any higher than those for autoignition. The only way that the converse could be true is either due to natural data scatter, or if the equipment is so badly designed that the pilot actually interferes with ignition. Only two workers have presented ‘a few grams plunged into a furnace’ data for piloted Tig. The values are 260ºC from Patten and 220–240ºC from Masařík, giving an average of 245ºC, which can be taken as identical to 250ºC. The conclusion is that piloting does not make any difference on Tig in tests of this type. Considering next piloted ignition results from radiant heating tests, it is evident that none are available at heating conditions barely enough for ignition. The available results are typically for specimens 12–25 mm thick and exposed for only 10–60 minutes. Shoub’s data indicate that much longer times are needed for specimens of these thickness before minimum conditions are approached. On the basis that piloted values should not be lower than autoignition, Tig = 250ºC can be provisionally assigned also as the piloted ignition temperature for radiant tests. Thus, it is concluded that 250ºC is the best estimate of the ignition temperature irrespective of piloting and irrespective of type of test, provided that heating conditions are just barely enough for ignition.

At this point, it is important to observe the nature of the low-heat ignitions. Moran, Li, and Spearpoint [33] all describe the same phenomenon: ignition starts as a glowing ignition and flaming is seen later, if at all. By the way, the glowing ignition temperature must not be confused with the temperature of the glowing spot. In a glowing ignition, a glow begins at one spot and very quickly reaches red-hot conditions (over 600ºC). This high temperature is not the glowing ignition Tig; instead the latter must be determined either by a thermocouple reading just before a steep jump takes place or by a thermocouple on the same surface but away from the spot of initial glow. The glowing ignition phenomenon also serves to explain why no difference is seen between autoignition and piloted ignition results. If flaming is

74 preceded by glowing, then the glowing zone can serve as a high temperature pilot, if subsequently sufficient pyrolysates emerge to be ignitable as a flame. Parenthetically, unlike wood, materials that are not susceptible to glowing ignition (e.g., thermoplastics) show a substantially lower Tig in the Setchkin furnace for piloted than for autoignition conditions.

‘A few grams plunged into a furnace’ tests generally share two features: very small specimen size, and exposure to conditions where, apart from radiant heating, the specimen is convectively heated (by contrast, in radiant heating tests the convective stream is cooling the specimen). There is one test series where fairly sizeable specimens were plunged into a furnace, and that is Prince’s 1915 study [34]. His original study reported that ignition was attained for furnace temperatures of 180–200ºC. Thomas [12] later estimated specimen surface temperatures by modeling and concluded that surface temperatures at ignition were 30ºC higher than the furnace temperature. This correction, which arises due to self-heating, is negligible for tiny specimens and increases with increasing specimen size. Since Thomas’ corrected values are in the same range as the raw values from ‘few grams’ specimen tests, the conclusion is that there is no specimen size dependence, at least when testing under heating conditions barely sufficient to cause ignition.

Considering next piloted ignition at higher fluxes, only the radiant tests can be considered, since ‘a few grams plunged into a furnace’ tests are not normally run this way. When the heat flux is high enough (and there is no good guidance on this point yet!) wood specimens ignite simply in a flaming mode, without antecedent glowing. For simplicity, it is best to consider first those results which pertain to a direct-flaming mode. For such heat fluxes, Tig ≈ 300– 350ºC covers all, or nearly all results of Janssens, Atreya, Abu-Zaid, Fangrat and Moghtaderi. Akita’s value of 450ºC, obtained by calculation, appears to be wrong since he did obtain ignitions at a furnace temperature of 350ºC (and did not try lower temperatures). For his 1.8 mm thick specimens, self-heating would be minimal, so the actual ignition temperature appears to have been below 350ºC, making his measurements also consistent. Of modern workers with good equipment, only the results of Li and Drysdale are outside this range and these are about 50ºC higher, for unknown reasons. Janssens [27] noted that the range can be further shrunk by considering the slight but systematic effect of wood type. His results for oven-dried specimens were: hardwoods 300–311ºC; softwoods 349–364ºC. At fluxes high enough to ensure a direct-flaming ignition, these values can be adopted for piloted Tig. Wood is comprised of three primary constituents—cellulose, hemicellulose, and lignin. Hemicellulose ignites at the lowest temperature, cellulose higher, and lignin higher yet [35]. Compared to hardwoods, softwoods have a smaller fraction of hemicellulose and a higher fraction of lignin, thus accounting for their higher Tig.

Next the intermediate-flux regime must be considered, where the heat flux is higher than the minimum flux, but is low enough for ignitions to be of the glowing → flaming type. These reported data span a sizeable range of 332–497ºC. Part of the scatter is probably due to experimental difficulties, since Urbas and Parker observed [36] that considerable care needs to be exercised to instrument properly a surface that is undergoing charring. Part of the difference, however, is real and is attributable to changed exposure conditions. Moran’s data are instructive here. Although intermediate data were scattered, as the flux was raised from 25 kW m-2 to 29 kW m-2, the ignition temperature rose from 255ºC to 301ºC while the ignition time dropped by 33%. The reason for the dependence of Tig on flux in this regime will be considered in the next section. The 300ºC value is significant, since wood pyrolysis involves competing mechanisms, with temperatures under 300ºC leading largely to charring, while

75 over 300ºC gasification being favored [37]. Thus, if heating conditions are such that the material does not exceed 300ºC, a glowing ignition is favored.

Concerning other systematic effects, at the minimum flux condition, Moran found no difference in Tig between oven-dried and room conditioned specimens. In the medium flux regime under piloted conditions, Janssens [38] concluded that Tig rises by 2ºC for each percent of moisture content increase. This will normally be insignificant for practical moisture contents. Specimen orientation (i.e., along-grain versus end-grain exposure) may also have an effect on Tig, but good enough data are not available to explore the issue. Almost all existing experimental data deal with along-grain exposures, which are also common in accidental fires.

GLOWING IGNITION MODELING A glowing ignition involves the direct surface oxidation of a material (heterogeneous reaction), thus Baer and Ryan [39] suggested that the simplest model for this is: ∂ λ ∂ 2 T = T ∂t ρ C ∂x 2 with the boundary condition: ∂T − λ = q′′ + B Q exp(−E / RT) ∂ &e s s s x x=0 ′′ where T = temperature, t = time, λ = thermal conductivity, ρ = density, C = heat capacity, q&e = irradiance, Bs = pre-exponential factor, Qs = heat of reaction, Es = activation energy for surface reaction, and R = universal gas constant. Based on this, Lengellé et al. [40] then showed that a solution for the ignition temperature Tig is: −1 E   B Q  T = s ln s s  ig   α ′′  R   q&e  where α = non-dimensional temperature rise associated with ignition. The α factor serves as an ignition criterion and they found empirically that α ≈ 0.15 corresponds to ignition. The equation shows that Tig decreases with decreasing irradiance, and Lengellé demonstrated that this indeed occurs experimentally for a number of propellants. Propellants are, of course, substances very different from wood, but Moussa et al. [41] proposed that the same equation be used in describing char oxidation occurring during smoldering of wood; however, they did not provide quantitative values for the kinetic constants. Fredlund [42] used a slightly different term in his model of wood combustion, but provided no experimental verification in the glowing ignition regime. Ohlemiller [43] noted that describing char oxidation of wood is difficult, since the char is not a unique chemical entity, but rather, is a substance whose characteristics are history-dependent. For a similar material, coal char [44], the chemical properties are, in fact, strongly dependent on the physical nature (pore structure) of the char that has been created, and it might be expected that this would also be important for wood. More complex heterogeneous reaction models that include pore-structure effects (and the possibility of both kinetically-limited and diffusion-limited reaction rates within these structures) are available for coal-char combustion [45], but such models have yet to be applied towards representing the ignition of wood. The above observations help to place in context the long times required for glowing ignition of wood—plywood required over 5 h in Shoub and Bender’s experiment. This long time period is associated with creating of a reactive porous char. The conclusion, thus, has to be that only qualitative rudiments are known for glowing ignition, and that quantitative modeling is not yet possible, largely because of an absence of experimental data.

76 FLAMING IGNITION FROM RADIANT HEATING Theory There is more than half a century of history in the development of both comprehensive and ‘practical’ theories of flaming ignition of wood materials. Janssens [27][38] reviewed them extensively and here only the salient feature will be reprised: his recommended method for plotting experimental data so that sound interpolations and extrapolations may be possible. His study, which was based on numeric approximations to an inert-body model of an igniting solid, entails plotting the ignition time raised to the –0.55 power on the y-axis and the external imposed heat flux (irradiance) on the x-axis. This is illustrated with Janssens’ own data in Figure 1. Since a straight line can be obtained when the data are plotted in this way, only two parameters are needed to describe the data fit. An obvious one to choose is the x- ′′ axis intercept, denoted as q&cr . The slope is a very small number, so it is more convenient to select the inverse of the slope and to designate it as Big. Thus, the equation describing the data plot is: −0.55 = []′′ − ′′ tig q&e q&cr / Big ′′ -2 -2 +0.55 In the example, q&cr = 9.3 kW m , Big = 201 kW m s . In general, it is found that ignition ′′ may not be possible at fluxes just slightly greater than q&cr , and a higher heat flux is necessary ′′ for ignition to actually occur. This latter value is designated q&min , the minimum flux for ignition. Thus, apart from the two parameters needed to describe the straight line, a third parameter is needed which denotes the lowest point on the line that has physical meaningfulness. Janssens presented a second method for thermally thin materials. Physically, whole wood is rarely used free-standing in minuscule thicknesses (e.g., < 1 mm), thus Janssens’ second procedure will not be presented here. But the ‘thermal thickness’ is not necessarily the same as the physical thickness, and substances of finite thickness which behave as thermally-thick bodies when initially heated will eventually respond as thermally- thin, if sufficient time has elapsed. This point is treated in the next section.

Janssens’ theory was mainly intended as an aid to using experimental data and was not intended to encompass all relevant physicochemical phenomena. Indeed, since it is an inert- solid theory, events in the gas phase are ignored and ignition is assumed to uniquely occur at the moment a certain face temperature is first attained. Much more refined theories have been put forth in recent years, for example, Yuen’s [46]. These have the limitations that they (a) require a large amount of input data, much of which may be unavailable or uncertain; and (b) difficult numeric computations must be performed for each problem; consequently, they are not useful as ‘data plotting aids.’ While advanced theories attempt to capture gas-phase ignition events, there is currently no theory available, simple or complex, which encompasses the possibility that a specimen may exhibit glowing ignition, glowing → flaming (2-step) ignition, or a direct-flaming ignition. Experimental results on piloted ignition From both theory and experiments, it is evident that a number of variables can affect the ignition time of thermally-thick, solid wood, of which density, thermal conductivity, moisture content, and geometric factors are probably the most important. Taking the last first, in testing, geometric effects show up as apparatus dependent factors, since no physical test rig can capture apparatus-independent properties of a material. Size of specimen is a geometric variable to consider, but Long et al. [47] noted that the only scale-dependent term in basic ignition theory is the convective heat transfer coefficient, hc, which varies with size L

77 ∝ 1 according to hc . The effect on ignition time is much smaller than the change in hc, L1/ 4 since heat losses are dominated by radiation, and would be negligible for all except huge changes of scale. But basic ignition theory does not deal with events in the gas phase and these may also have an effect. Within a single test apparatus, experimental data suggest that the size effect is very small [48], although when comparisons are made where both the scale and the basic apparatus are changed, somewhat larger differences crop up [49][50]. In any case, currently there is a sizeable database of test results only from the Cone Calorimeter, so for consistency, only Cone Calorimeter data obtained on samples exposed in the along-grain orientation will be considered. According to basic theory, for thermally-thick materials ∝ λρ ()− 2 tig C Tig To . Since radiant ignition data are easy to obtain, but become much more difficult if surface temperatures need to be accurately measured, the consequence is that most investigators record only the flux and the ignition time. Consequently, it is best to treat Tig ∝ λρ from such data sets as part of the unknown constants to be fitted, thus taking tig C . Now, values of heat capacity, C, tend to vary little among members of a chemical family, and this appears to also be a reasonable conclusion for woods. Density, however, can vary by about a factor of 10, if exotic woods are included. Thermal conductivity increases with increasing temperature, with the simplest assumption being that λ ∝ ρ n , where the value of n remains to ∝ ρ m be determined. Thus, it seems appropriate to seek a correlation where tig , where m is also to be determined. This is not a novel idea, and Hallman [51] took a similar approach some 30 years ago. Moisture content can have a complex effect, both because it directly affects the thermophysical properties, and because, if it were to be treated accurately, an inert- substance model is no longer a viable starting point for a theoretical treatment. To make an accurate treatment of moisture, the extremes of green wood to oven-dried wood would have to be considered. Green wood can have MC > 100%, but there are no available ignition data on it, with the literature containing data only for oven-dried specimens and ones that are equilibrated to room conditions. For room-conditioned wood specimens, MC depends on the humidity present, but across the US it normally spans only the range of 4–14% [52], which is a small range and only covers the ‘zero-end’ of the scale. Most test results available are either for the oven-dried condition or for 9–12% moisture content, obtained by room-conditioning the specimens. As indicated above, Janssens concluded that moisture slightly increases Tig, but this can be ignored unless the wood is green (for which no data are available, anyway).

To find a correlation, a large number of published [31][33][53][54][55][56] and unpublished [57][58][59] data sets were collected. These covered four test conditions: oven-dried horizontal, oven-dried vertical, room-conditioned horizontal, and room-conditioned vertical. Figure 2 shows the results for oven-dried horizontal specimens [56][58]; with one data set [31] not used due to excessive outliers. The densities spanned 170–850 kg m-3. Since this data set showed a relatively tight correlation, the exponent for the density term was derived from the data fit on this data set and fixed at that value for the remaining data fits. The value = −0.55 ρ −0.4 plotted on the y-axis is Y, which was taken as Y tig . The other data sets showed higher scatter, for example, Figure 3. Table 3 gives a summary of the correlations obtained. Figure 4 shows that the correlations are very similar and that it is reasonable to assign an ‘overall’ correlation. Clearly a dry specimen ignites quicker than a moist one, but this is somewhat violated in the correlations, and this is one reason why it is best to assign a single correlation, with the realization that moisture effects are swamped by general data scatter. It is also known that vertically-oriented specimens take longer to ignite than do horizontally- oriented ones [54], but again the scatter of the data does not permit this to emerge from the

78 correlations. Based on these considerations, the estimating rule for radiant heating ignition of wood becomes: 130ρ 0.73 t = ig ()′′ − 1.82 q&e 11.0 Table 3 Summary of data correlations for piloted radiant ignition in the Cone Calorimeter ′′ Conditions q&cr Const. Tot. data Data points points used horiz., 0% MC 9.8 159 31 26 horiz., room 12.2 128 103 94 vert., 0% MC 11.5 99 67 48 vert., room 9.0 133 53 48 overall 11.0 130

According to theory, it would appear that the exponent for ρ is unusually low, but the reason for this is not clear. The root-mean-square error of the predictions is 64%, which indicates that predicting times to ignition can only be done semi-quantitatively, but this must also be placed in the context that experimental data went from 2.5 to 4200 s, or a range of 1 : 1680. A close inspection of Figure 3 also reveals that below about 15 kW m-2, the points deviate systematically above the straight line. This is as might be expected, since the theory is based on a thermally thick material, and wood specimens 12–25 mm thick no longer behave in a thermally thick manner when heated for a long time. It is possible to eliminate this systematic bias by fitting exponents higher than 1.82 to the irradiance factor, for example, 2.8 as suggested by Wesson [60]. But overall scatter still remains large and the treatment becomes wholly empirical.

The minimum flux for ignition is often the quantity of interest. In 1965, McGuire [61] suggested that this value can be taken as ca. 12.5 kW m-2 for most wood materials apart from low-density fiberboard. A value of 12.5 kW m-2 has subsequently been used for design purposes in many countries. This is indeed the value that is customarily obtained in the Cone Calorimeter and in other test methods where the time allotted for observation of ignition is 10–20 minutes. But lower values have been found, although not widely publicized. Spearpoint [33] recently explored both low-flux ignition and end-grain ignition of woods. Almost all ignition results available for wood are performed on specimens oriented towards the heat source along the grain, but different results are obtained when exposed to the end- ′′ -2 grain. For along-grain exposures, Spearpoint found q&min = 12.5 kW m for redwood and somewhat less than 12 kW m-2 for maple. But for end-grain ignition of maple, the lowest flux -2 -2 ′′ at which ignition occurred was 8 kW m , with no ignition at 7 kW m , making q&min = 7.5 kW m-2. The minimum flux for end-grain ignition of redwood was not fully explored, but was found to be below 9 kW m-2. For ignitions occurring at fluxes below 10 kW m-2, a glowing ignition preceded flaming. The times associated with the low-flux ignitions were notably long, it taking 2680 s for end-grain ignition of maple at 8 kW m-2, and 4200 s for along-grain -2 -2 ′′ ignition at 12 kW m . On this basis, one might conclude that 7.5 kW m is q&min for piloted ignition of wood, but the value for piloted ignition cannot be higher than for autoignition and the latter may be low indeed (see below).

Generally, ‘piloted ignition’ means the presence of a flame or a spark in the gas phase where pyrolysates accumulate. But it is also possible to apply a gas flame directly onto a surface as an ‘impinging pilot,’ in which case much less radiant heating is needed to achieve ignition

79 ′′ since a local heat flux concentration is created. An old FRS study [62] showed q&min = 5 kW m-2 for Western red cedar and Douglas fir. No other published studies exist. Apart from surface-applied pilots, both the type of pilot and its location can affect ignition times. Several studies [54][63] produced limited data—more studies would be needed to quantify trends reliably. It is also possible to heat a wood surface by applying a relatively-uniform ‘wall of flame’ onto it, and this is discussed later. Experimental results on autoignition Unlike piloted ignition, autoignition of wood under radiant heating conditions has been studied by only a few researchers, most notably Simms, who conducted various experiments at FRS in the 1950s and ’60s. In a 1952 study, he tested 6 different species of wood using 19 mm thick specimens [64]. The results, including the correction for a 20% flux mis-calibration [65], are shown in Table 4. In a 1961 study [66], he reported an enormous value of up to 117 kW m-2 for autoignition of blackened oak and cedar specimens. In a 1967 study [67], he reported minimum fluxes for piloted ignition that were similar to the corrected 1952 values, but autoignition values reported were quite a bit higher, being ca. 40–50 kW m-2. In his 1961 study, Simms noted that a draft strong enough to be turbulent was helpful in reducing the ′′ q&min . This was evidently a gas-phase effect, but even today there is no systematic knowledge on gas-phase ignition effects. In another study [68], Simms concluded that the quantitative effect of the rather small exposure size of 8 mm is nearly negligible, so presumably the ′′ enormous q&min values in the 1961 study were mainly due to insufficiently long test time. Table 4 Minimum flux for autoignition of wood, as reported by various researchers ′′ Study Orient. MC Draft Specimen Max. time q&min Notes (%) size exposed of test (kW m-2) Lawson, Simms V 0 N 50 × 50 mm 20 min 29–33 (1952) Simms (1961) V 0 Y 8 mm ø 14 s 75–100 blackened N 18 s 117 surface Simms, Law V 0 N 76 × 76 mm 70 s 46 (1967) 150 × 150 mm 79 s 42 Moran V 0 Y 50 × 50 mm 9 min 25 Shields et al. H ≈10 N 100 × 100 mm 96 s 30–40 V 59 s 40–50 H ≈10 N 165 × 165 mm 12 min < 20 ISO 5657 test Shoub, Bender V ≈10 N 920 × 920 mm 3.9–5.2 h 4.3

Moran [15] examined the ignition of vertical panels of 6.4 mm thick ponderosa pine using an ′′ -2 electric radiant panel and found q&min = 25 kW m . Shields et al. [54] examined the autoignition of Sitka spruce in the Cone Calorimeter and in the ISO 5657 apparatus. They exposed specimens in increments of 10 kW m-2, so their results were only approximate. Since the heater arrangements have some similarity, it is not clear why the values obtained in the Cone Calorimeter and the ISO 5657 apparatus were not closer. Shields’ data does illustrate that it is much more difficult to achieve autoignition in the vertical orientation than in the horizontal orientation. The above studies were all of less than 20 minutes duration. Only the study, by Shoub and Bender [18] involved longer-term exposures. They used an electric radiant panel operating at an effective black-body face temperature of 273ºC and producing a heat flux of 4.3 kW m-2 at the center of the specimen, and lower heat fluxes at the edges. While they did not test any whole woods, they tested 13 mm plywood. It ignited at the 4.3 kW m-2 flux, but required waiting over 5 hours. In their tests, they also documented that the

80 face temperatures of the specimens in some cases reached temperatures higher than that of the radiant source, indicating that self-heating of the material was important and that assuming an inert solid would not be appropriate. It should be of high priority that modern- day researchers attempt to repeat these experiments and verify their results. The conclusion— pending a verification of Shoub and Bender’s results—is that wood will autoignite at about 4.3 kW m-2, if exposed for hours, rather than minutes. For short-term exposures, a value of 20 kW m-2 perhaps best captures the research results.

At any given irradiance, if ignition occurs under both autoignition and piloted ignition conditions, it is evident that ignition times for the latter will be shorter (unless the pilot is badly placed). A tractable theory, such as Janssens’, models only the solid phase, so the presumed conclusion would be that ignition times do not change. A more refined point of view would be to assume that for autoignition, heating up the solid to the same temperature suffices as for the piloted case, but that afterwards a delay time must be added to account for gas-phase events. A theory of this sort has not been developed, however. Experimentally, even though there is a great deal of scatter (Figure 5), the results of Shields et al. [54] can be used to derive an equation: ()− ′′ ⋅ tig (autoignition) = 2.86 0.0172 q&e tig (spark) Thus, for example, at a flux of 25 kW m-2, under autoignition, ignition times can be expected to be 2.43× those for the spark-ignition case, while at 50 kW m-2 the factor drops down to 2.0×. Since the highest experimental flux was 70 kW m-2, the rule should not be extrapolated to greatly higher fluxes. Also, due to the data scatter, the guidance is only semi-quantitative.

IGNITION FROM MISCELLANEOUS HEAT SOURCES There is very little data on ignition of wood from flames, despite the fact that this is how we light our fireplaces. When a thin piece of wood is lit at the bottom, burning may continue to completion. But a thick piece of wood will not undergo self-sustained combustion under the same circumstances. Bryan [69] reports that the maximum thickness for self-sustained burning, given a flaming ignition at the bottom of a vertical piece, is about 19 mm. In a horizontal orientation, even 12 mm thick specimens have been found too thick for self- sustained burning [70]. Using the methenamine pill test (a standard test for floor coverings), it was found [71] that no ignition occurs for any of a wide variety of wood products tested in thicknesses of 10 – 21 mm. Ignitability of wood boards has also been examined [72] using the ISO 11925-2 small-burner test. Using a 30 s flame exposure to the surface, ignition rarely occurred and never spread to the 150 mm limits, even with specimens as thin as 2 mm. For 30 s bottom-edge impingement, specimens of 18 mm thickness or less commonly ignited, but only ones of 10 mm thickness or less generally reached the 150 mm mark.

Ebeling and Welker [73] studied the ignition of wood panels when exposed to a flame, with the flame being applied against the whole face. They tested oak, white pine, redwood, and yellow pine, with the results giving the correlation: = ρ 0.94 ()′′ −1.82 tig 41.3 q&e although there was a wide spread of results. The above equation implies that the critical flux is identically zero, which is at least partly due to the fact that there is no convective cooling of the surface in a flame-ignition test. For the same reason, their flame ignition times were a fair bit shorter than times obtained by applying the same heat flux in a radiant heating test.

Even though convective heating is an important feature in a Setchkin-type apparatus, there has been no scientific study where ignition would be primarily from convective heating.

81

When a sufficiently high voltage is impressed across a tree or a wood member, an arc tracking process takes places. Wood first dries out at the electrodes, then a carbonized channel starts to form. Given enough time and voltage, sufficient heating of the carbonized track takes place that the electric current passing through the track heats up the wood to ignition. This process has been studied by several researchers [74][75][76][77]. It is mainly of concern in connection with high voltage wiring, including power line poles and neon signs installed near wood surfaces.

Ignition with laser radiation produces very different results than radiation from flames or grey-body radiators, for reasons not fully explained. Kashiwagi [78] exposed horizontally- ′′ oriented red oak specimens to laser radiation at 10.6 µm and found high values of q&min ≈ 80 kW m-2 for autoignition and 55 kW m-2 for piloted ignition. This is also common with laser ignition of other substances, e.g., plastics. Ignition from nuclear weapons has been simulated [79] by use of radiant exposure from an arc-image furnace. The results for the brief, high- intensity pulses were expressed in terms of energy fluence. Using 13 mm thick Douglas fir, transient flaming was observed for an energy fluence of 1090 kJ m-2 (480 kt bomb) and sustained flaming at 1300 kJ m-2 (1180 kt) or higher. Yellow poplar of 1.6 mm thickness also showed sustained flaming at 1090 kJ m-2, but no transient flaming regime.

IGNITION FROM HOT BODIES, FIREBRANDS, AND SMOLDERING Glowing and smoldering are similar, but not identical mechanisms of ignition. Smoldering is, by definition, a self-sustained process. Ignition and consumption of a wood material by glowing, on the other hand, can occur if it is subject to sufficient radiant or convective heating, without a requirement that the process continue, should the external heat source be removed. Firebrands themselves may be flaming or glowing, and they may, in some cases, initially cause flaming in the target fuel, although a smoldering ignition is the usual concern.

Self-sustained smoldering occurs easily in various wood products which are highly porous or finely divided (fiberboard, wood shavings, rotted wood, etc.). Whole wood, however, is only slightly porous to the inflow of oxygen and will not smolder as a single surface facing open air. Ohlemiller [80] reports that by supplying external heating at ca. 10 kW m-2, wood can be made to burn in a glowing mode; this of course is not smoldering, since it is not self- sustained. By preheating the bulk of the wood sufficiently, continued combustion can be maintained. This can be seen in a fireplace where individual pieces may continue glowing even after a ‘three-log’ effect no longer exists. Only a limited number of experimental studies exist on the question of minimum conditions necessary to start wood smoldering. Ohlemiller [81] conducted experiments where smoldering was achieved by providing a ‘three-log’ arrangement and igniting the surfaces with flat electric heaters. Even with the optimal geometry, air flow velocity had to be within a close range for sustained smoldering to be seen.

Solid wood is most commonly ignited by firebrands during wildland fires. Humidity plays a strong role in the process, and wildland fires often involve extremes of high temperature, low humidity, and strong wind gusts. Only a few laboratory studies have been conducted on the ignitability of solid wood by firebrands. CSIRO researchers [82][83] found that some surprisingly small (0.8–12 g) wood cribs sufficed for ignition. An inside-corner (‘re-entrant corner’) geometry of the siding was especially conducive to ignition. Hamada [84] found that no-wind conditions, red-hot brands of about 5 mm diameter caused ignition, but in an 8 m s-1 wind, even brands of 2.5 mm were likely to cause ignition. Low RH values (20%) were

82 needed for this to occur. Applying flames to the surface of a wood structural member will not result in smoldering ignition, unless the flame is applied for so long that the wood member is largely burned up. Specifically, it has been demonstrated [85] that applying the flame from an acetylene/air plumber’s torch directly onto wood studs for periods of 1–5 minutes leads to local charring but no sustained combustion of any type once the torch is removed and the flames self-extinguish.

A special problem is one where ignition of wood occurs from steam pipes or from a metal heating system part which is in contact with the wood for a long time. Under long-term heating (months-to-years), it appears that wood can ignite when a surface is held at a temperature lower than the ignition temperature determined from tests that last a short time (minutes-to-days). The information largely comes from case histories and good experiments are lacking. Babrauskas [86] recently reviewed the state of the art on this topic.

CONCLUSIONS Some aspects of the wood ignition problem are well-known, and these can be used in routine engineering applications. This is primarily true of ignition times for piloted ignition, provided that fluxes too close to the minimum are not considered. But despite more than a century of scientific research, many other aspects of wood ignition are poorly known. Foremost is a lack of study of ignition at minimum-flux conditions, including an understanding of glowing ignition. A simple theory exists for glowing ignition, but it cannot be used without good experimental data, and detailed, reliable experimental data are lacking. Part of the problem is that only a few experimentalists report visual observations along with their numeric data. Analysis of available data leads to the summary given in Table 5 and schematically depicted in Figure 6. High fluxes (e.g., over 80 kW m-2) are not listed since there is little information, but also this regime is of less interest to fire safety. Most researchers have conducted much too short tests in attempting to define ‘minimum’ conditions, thus only a single study is the basis for observing that ignition may occur at heat fluxes as low as 4.3 kW m-2. It is, of course, likely that there is not a unique minimum flux value, but that various factors—apart from inadequate duration of experiments—can affect its value. Also needing to be quantified is the flux value dividing the medium flux (flaming ignition) from low flux (ignition starts with glowing) regimes. Somewhat related to the lack of knowledge about glowing ignitions is the lack of knowledge on ignitions from hot bodies. Experimental data on this topic are so scarce that it can only be concluded that ignitions are possible under some surprisingly mild attacks, e.g., firebrands of a few grams. Ignition of wood in actual fires often is due to direct flame contact with the material, but again guidance on this topic is minimal.

Table 5 Summary of ignition temperature results Flux Minimum Low Medium Ignition type glowing or glowing/flaming flaming Tig (ºC), piloted 250 350 – 400 peak, lower for 300 – 310 hardwoods fluxes close to minimum. 350 – 365 softwoods Tig (ºC), autoignition 250 no data 380 – 500 ??

83 REFERENCES

1. Standard Test Method for Determining Ignition Temperature of Plastics (ASTM D 1929), American Society for Testing and Materials, West Conshohocken PA. 2. Hill, H. B., On the Behavior of Sound and Decayed Wood at High Temperatures, Proc. Amer. Acad. Arts and Sciences 22, 482-492 (1887). 3. Bixel, Edward C., and Moore, Howard J., Are Fires Caused by Steam Pipes? (B. S. thesis), Case School of Applied Science, Pittsburgh (1910). 4. Banfield, W. O., and Peck, W. S., The Effect of Chemicals on the Ignition Temperature of Wood, Canadian Chemistry and Metallurgy 6, 172-176 (Aug. 1922). 5. Brown, C. R., The Determination of the Ignition Temperatures of Solid Materials (D.Sc. thesis), The Catholic University of America, Washington (1934). 6. VanKleeck, A., A Preliminary Study of Ignition Temperatures of Finely Chopped Wood (Project L- 179), Forest Products Lab., Madison WI (1936). 7. As cited in: Wood and Wood-Based Products, J. M. Cholin, ed., Fire Protection Handbook, A. E. Cote, and J. L. Linville, eds., 18th ed., National Fire Protection Assn., Quincy MA (1997). 8. Graf, S. H., Ignition Temperatures of Various Papers, Woods, and Fabrics (Oregon State College Bull. 26), Oregon State College, Corvallis (March 1949). 9. Angell, H. W., Gottschalk, F. W., and McFarland, W. A., Ignition Temperature of Fireproofed Wood, Untreated Sound Wood and Untreated Decayed Wood, British Columbia Lumberman 33, 57-58, 70-72 (Sept. 1949). 10. Fons, W. L., Heating and Ignition of Small Wood Cylinders, Ind. and Eng. Chem. 42, 2130-2133 (1942). 11. Narayanamurti, D., A Note on Pyrolysis and Ignition of Wood, Current Science 27, 22-23 (1956). 12. Thomas, P. H., Simms, D. L., and Theobald, C. R., The Interpretation of Some Experimental Data on the Ignition of Wood (Fire Research Note No. 411), Fire Research Station, Borehamwood, UK (1959). 13. Akita, K., Studies on the Mechanism of Ignition of Wood, Report of the Fire Research Institute of Japan 9, 1-44, 51-54, 77-83, 99-105 (1959). 14. Simms, D. L., Ignition of Cellulosic Materials by Radiation, Combustion and Flame 4, 293-300 (1960). 15. Moran, H. E., jr., Effectiveness of Water Mists for Protection from Radiant Heat Ignition (NRL Report 5439), US Naval Research Laboratory, Washington (1960). 16. Patten, G. A., Ignition Temperatures of Plastics, Modern Plastics 38, 119-122, 180 (July 1961). 17. Buschman, A. J., Ignition of Some Woods Exposed to Low Level Thermal Radiation (NBS Report 7306), [U.S.] Natl. Bur. Stand., Washington (1961). 18. Shoub, H., and Bender, E. W., Radiant Ignition of Wall Finish Materials in a Small Home (NBS 8172), [U.S.] Natl. Bur. Stand, Washington (1964). 19. Tinney, E. R., The Combustion of Wooden Dowels in Heated Air, pp. 925-930 in 10th Symp. (Intl.) on Combustion, The Combustion Institute, Pittsburgh (1964). 20. Simms, D.L., and Law, M., The Ignition of Wet and Dry Wood by Radiation, Combustion and Flame 11, 377-388 (1967). 21. Melinek, S. J., Ignition Behaviour of Heated Wood Surfaces (FR Note 755), Fire Research Station, Borehamwood, UK (1969). 22. Jach, W., Das Verhalten von Holz und Holzwerkstoffen bei Dauereinwirkung von Temperaturen unterhalb des Flamm- und Brennpunktes, Mitteilungen der deutschen Gesellschaft für Holzforschung, Nr. 56, 12-17 (1969). 23. Smith, W. K., and King, J. B., Surface Temperatures of Materials during Radiant Heating to Ignition, J. Fire & Flammability 1, 272-288 (1970). 24. Atreya, A., Pyrolysis, Ignition and Fire Spread on Horizontal Surfaces of Wood (Ph.D. dissertation), Harvard University, Cambridge MA (1983). 25. Atreya, A., Carpentier, C., and Harkleroad, M., Effect of Sample Orientation on Piloted Ignition and Flame Spread, pp. 97-109 in Fire Safety Science — Proc. 1st Intl. Symp., Hemisphere, Washington (1986). 26. Abu-Zaid, M., Effect of Water on Ignition of Cellulosic Materials (Ph.D. thesis), Michigan State Univ., East Lansing MI (1988). 27. Janssens, M. L., Fundamental Thermophysical Characteristics of Wood and Their Role in Enclosure Fire Growth (Ph. D. dissertation), University of Gent, Belgium (1991).

84

28. Li, Y., and Drysdale, D., Measurement of the Ignition Temperature of Wood, pp. 380-385 in Fire Science and Technology—Proc. First Asian Conf., Intl. Academic Publishers, Beijing (1992). 29. Masařík, I., Ignitability and Burning of Plastic Materials: Testing and Research, pp. 567-577 in Interflam ’93, Interscience Communications Ltd., London (1993). 30. Fangrat, J., Hasemi, Y., Yoshida, M., and Hirata, T., Surface Temperature at Ignition of Wooden Based Slabs, Fire Safety J. 27, 249-259 (1996); 28, 379-380 (1997). 31. Moghtaderi, B., Novozhilov, V., Fletcher, D. F., and Kent, J. H., A New Correlation for Bench-scale Piloted Ignition Data of Wood, Fire Safety J. 29, 41-59 (1997). 32. Standard Test Method for Heat and Visible Smoke Release Rates for Materials and Products using an Oxygen Consumption Calorimeter (E 1354), American Society for Testing and Materials, West Conshohocken PA. 33. Spearpoint, M. J., Predicting the Ignition and Burning Rate of Wood in the Cone Calorimeter Using an Integral Model (NIST GCR 99-775), Nat. Inst. Stand. and Technol., Gaithersburg MD (1999). 34. Prince, R. E., Tests on the Inflammability of Untreated Wood and of Wood Treated with Fire- Retarding Compounds, Proc. NFPA. 19, 108-158 (1915). 35. Buchanan, M. A., The Ignition Temperature of Certain Pulps and Other Wood Components, TAPPI, 35, 209-211 (1952). 36. Urbas, J., and Parker, W. J., Surface Temperature Measurements on Burning Wood Specimens in the Cone Calorimeter and the Effect of Grain Orientation, Fire and Materials 17, 205-208 (1993). 37. Shafizadeh, F., Utilization of Biomass by Pyrolytic Methods, pp. 191-199 in Proc. TAPPI 1997 Joint Forest Biology/Wood Chemistry Mtg., Madison WI (1977). 38. Janssens, M., Piloted Ignition of Wood: A Review, Fire and Materials 15, 151-167 (1991). 39. Baer, A. D., and Ryan, N. W., Ignition of Composite Propellants by Low Radiant Fluxes, AIAA J. 3, 884-889 (1965). 40. Lengellé, G., Bizot, A., Duterque, J., and Amiot, J.-C., Ignition of Solid Propellants, La Recherche Aérospatiale—English Edition, No.2, 1-20 (1991). 41. Moussa, N. A., Toong, T. Y., and Garris, C. A., Mechanism of Smoldering of Cellulosic Materials, pp. 1447-1457 in 16th Symp. (Intl.) on Combustion, The Combustion Institute, Pittsburgh (1976). 42. Fredlund, B., A Model for Heat and Mass Transfer in Timber Structures during Fire (Report LUTVDG/TVGG-1003), Dept. of Fire Safety Engineering, Lund University, Lund, Sweden (1988). 43. Ohlemiller, T. J., Modeling of Smoldering Combustion Propagation, Prog. Energy & Comb. Sci. 11, 277-310 (1985). 44. Khitrin, L. N., Physics of Combustion and Explosion, Israel Program for Scientific Translations, Jerusalem (1962). 45. Smith, I. W., The Combustion Rates of Coal Chars: A Review, pp. 1045-1065 in 19th Symp. (Intl.) on Combustion, The Combustion Institute, Pittsburgh (1982). 46. Yuen, R. K.-K., Pyrolysis and Combustion of Wood in a Cone Calorimeter (Ph.D. dissertation), The University of New South Wales, Australia (1998). 47. Long, R. T. jr., Torero, J. L., Quintiere, J. G., and Fernandez-Pello, A. C., Scale and Transport Considerations on Piloted Ignition of PMMA, pp. 567-578 in Fire Safety Science—Proc. 6th Intl. Symp., Intl. Assn. for Fire Safety Science (2000). 48. Hu, X.-F., and Clark, F. R. S., The Use of the ISO/TC92 Test for Ignitability Assessment, Fire and Materials 12, 1-5 (1988). 49 . Mikkola, E., Ignitability Comparisons between the ISO Ignitability Test and the Cone Calorimeter, J. Fire Sciences 9, 276-284 (1991). 50. Nussbaum, R. M., and Östman, B. A.-L., Larger Specimens for Determining Rate of Heat Release in the Cone Calorimeter, Fire and Materials 10, 151-160 (1986). 51. Hallman, J. R., Ignition Characteristics of Plastics and Rubber (Ph.D. Dissertation), University of Oklahoma, Norman (1971). 52. Peck, E. C., Moisture Content of Wood in Dwellings (Circular No. 239), US Dept. of Agriculture, Washington (1932). 53. Henderson, A., Predicting Ignition Time Under Transient Heat Flux Using Results from Constant Heat Flux Experiments (Fire Engineering Research Report 98/4), School of Engineering, University of Canterbury, New Zealand (1999). 54. Shields, T. J., Silcock, G. W., and Murray, J. J., The Effects of Geometry and Ignition Mode on Ignition Times Obtained using a Cone Calorimeter, Fire and Materials 17, 25-32 (1993). 55. Mikkola, E., Puupinnan syttyminen (Research Notes 1087), Valtion Teknillinen Tutkimuskeskus, Espoo, Finland (1989).

85

56. Dlugogorski, B., Pope, D. M., Moghtaderi, B., Kennedy, E. M., and Lucas, J. A., A Study on Fire Properties of Australian Eucalyptus, pp. 57-72 in Wood & Fire Safety—4th Intl. Scientific Conf., The High Tatras, Slovak Republic (2000). 57. Janssens, M. L., private communication (2001). 58. Babrauskas, V., unpublished NIST test results. 59. Grexa, O., private communication (2001). 60. Wesson, H. R., The Piloted Ignition of Wood by Radiant Heat (D.Eng. dissertation), Univ. Oklahoma, Norman (1970). 61. McGuire, J. H., Fire and the Spatial Separation of Buildings, Fire Technology 1, 278-287 (1965). 62. Pickard, R. W., Simms, D. L., and Walters, J. E. L., The Ignition by Radiation of Wood Protected by Some Common Paints (FPE Note No. 61/1951), Fire Research Station, Borehamwood (1951). 63. Simms, D. L., and Hird, D., On the Pilot Ignition of Materials by Radiation (FR Note 365), Fire Research Station, Borehamwood, England (1958). 64. Lawson, D. I., and Simms, D. L., The Ignition of Wood by Radiation, Brit. J. Appl. Phys. 3, 288-292 (1952). 65. Simms, D. L., On the Pilot Ignition of Wood By Radiation, Combustion and Flame 7, 253-261 (1963). 66. Simms, D. L., Experiments on the Ignition of Cellulosic Materials by Thermal Radiation, Combustion and Flame 5, 369-375 (1961). 67. Simms, D.L., and Law, M., The Ignition of Wet and Dry Wood by Radiation, Combustion and Flame 11, 377-388 (1967). 68. Simms, D. L., Ignition of Cellulosic Materials by Radiation, Combustion and Flame 4, 293-300 (1960). 69. Bryan, J., The Fire Hazard, Wood 8, 260-262 (1943). 70. Wilson, J. A., A Different View on Plastics Fire Hazard Classification, NFPA Q. 55, 162-164 (Oct. 1962). 71. Tsantaridis, L., CEN Tablet Test Results for Wood Floorings (L-Rapport 9702009), Trätek, Stockholm (1997). 72. Tsantaridis, L., CEN Ignitability Test Results for Wood Building Products (L-Rapport 9702010), Trätek, Stockholm (1997). 73. Ebeling, K. E., and Welker, J. R., The Ignition of Wood by Direct Flame Contact (Report No. OURI- 1873-2), Univ. of Oklahoma Research Institute, Norman OK [1973]. 74. Kinbara, T., and Sue, S., On the Outbreak of Fire Due to Leakage of Electricity from Neon Transformer through Planking [in Japanese], Bull. Fire Prev. Soc. of Japan 2:2, 39-41 (1953). 75. Kinbara, T., and Takizawa, K., Ignition of a Salt-soaked Wooden Board by an Electric Current through It [in Japanese], Bull. Fire Prev. Soc. of Japan 11:2, 26-31 (Dec. 1961). 76. Sanderson, J. L., Carbon Tracking: Poor Insulation Combined with Contaminants is Potential Fire Cause, Fire Findings 8:3, 1-3 (2000). 77. Blackburn, T. R., and Pau, L. T., Characteristics of a Simulated High Impedance Fault, pp. 301-304 in Electric Energy Conf. 1985, Newcastle, Australia (1985). 78. Kashiwagi, T., Experimental Observation of Radiative Ignition Mechanisms, Combustion and Flame 34, 231-244 (1979). 79. Critical Radiant Exposures for Ignition of Tinder and Combustible Materials (Part I – Wood), Naval Applied Science Lab., Brooklyn NY (1965). 80. Ohlemiller, T. J., Smoldering Combustion (NBSIR 85-3294), [U.S.] Natl. Bur. Stand., Gaithersburg, MD (1986). 81. Ohlemiller, T. J., Smoldering Combustion Propagation on Solid Wood, pp. 565-574 in Fire Safety Science—Proc. 3rd Intl. Symp., Elsevier, New York (1991). 82. McArthur, N. A., and Lutton, P., Ignition of Exterior Building Details in Bushfires: An Experimental Study, Fire and Materials 15, 59-64 (1991). 83. Dowling, V. P., Ignition of Timber Bridges in Bushfires, Fire Safety J. 22, 145-168 (1994). 84. Hamada, M., et al., Experiments on the Ignition due to Fire Brands [in Japanese], Fire Research— Reports from the Fire Science Research Committee, Property and Casualty Insurance Rating Organization of Japan, Tokyo (1951). 85. Babrauskas, V., and Hall, J., unpublished test work (2000). 86. Babrauskas, V., Pyrophoric Carbon…The Jury is Still Out, Fire and Arson Investigator 51:2, 12-14 (Jan. 2001).

86 0.20 4

0.18

0.16 3

-0.55 0.14 t 0.12

Y 2 0.10

0.08 1 0.06 Transformed time, Transformed 0.04

0.02 0 0 10203040506070 0.00 Irradiance (kW m-2) 0 1020304050 -2 Irradiance (kW m ) Figure 2 Correlation for oven-dried horizontal specimens Figure 1 Janssens’ piloted ignition results for Blackbutt, oven-dried, vertical orientation

7 7 Vertical, room-cond. Horizontal, room-cond. Shorter 6 6 Vertical, 0% MC times Horizontal, 0% MC

5 5

4 4 Y Y 3 3 Longer times 2 2

1 1

0 0 0 102030405060708090100 0 102030405060708090100 -2 Irradiance (kW m-2) Irradiance (kW m )

Figure 3 Correlation for room-conditioned Figure 4 Data correlations obtained for the horizontal data; hollow points were not used to four data sets derive the correlation

87

3.0

2.5

2.0 Chipboard, H Plywood, H Spruce, H 1.5 Chipboard, V Plywood, H Spruce, V

Ratio (autoignition/spark ignition) time Data correlation 1.0 0 20406080 -2 Irradiance (kW m ) Figure 5 Ratio of ignition times, autoignition/spark ignition

250

Glowing or glowing/flaming

Ignition temperature (°C) temperature Ignition Flaming (softwoods) Flaming (hardwoods)

Irradiance (kW m-2 ) Figure 6 The effect of irradiance of piloted ignition temperature

88 THEORIES OF THE COMBUSTION OF WOOD AND ITS CONTROL

A Survey of the Literature

By

F. L. BROWNE, Chemist

Forest Products Laboratory,–1 Forest Service U. S. Department of Agriculture

Introduction

As a rule wood does not burn directly (44, 79, 114 ). –2 It first undergoes ther- mal degradation, or pyrolysis --some of the products of which are combus- tible gases, vapors, or mists. Under appropriate conditions the products may be set afire and, if enough of their heat of combustion is retained by the wood to maintain the pyrolysis,the burning may continue of its own accord until the wood has been consumed except for inorganic products left as ash. Ordinarily wood is set afire by bringing to bear enough heat to start active pyrolysis, and then applying a pilot flame or other source of high temperature to the combus- tible gaseous products after they have escaped and become mixed with air. In the absence of a pilot flame,much more heat must be supplied before the py- rolysis products will take fire spontaneously. The minimum rate of heating necessary for ignition by pilot flame is of the order of 0.3 calorie per square centimeter per second,but for spontaneous ignition it is of the order of 0.6 calorie per square centimeter per second (68 ).

Wood may be said to burn directly if its surface is irradiated so intensely that the temperature is raised to the point of spontaneous ignition within a fraction of a second,so that pyrolysis and combustion are practically simultaneous. Even then direct combustion is confined to a thin surface layer. Thin sheets of alpha cellulose that are exposed to short pulses of radiant power of 20 to 30 calories per square centimeter per second lose only part of their thickness in flame, leaving an extremely shallow layer of char that is backed by apparently unchanged cellulose (79). Direct combustion of explosive violence can occur

–Maintained1 at Madison,Wis.,in cooperation with the University of Wiscon- sin.

–Underlined2 numbers in parentheses refer to Literature Cited at the end of this report.

Rept. No. 2136 -1- when fine,dry particles of wood are distributed in air in such proportions that each particle is in contact with enough oxygen for its combustion and is close enough to its neighbors to ignite them quickly, in a manner analogous to a branching chain reaction (44 ).

The Course of Pyrolysis

The present interest is in normal combustion preceded by pyrolysis, When a piece of wood is heated out of contact with air,zones (44, 79, 84) develop parallel to the heat-absorbing surface,delimited by temperatures attained. The zones are well marked in wood because of its relatively low thermal con- ductivity (75) and density and relatively high specific heat (22 ).

Zone A, up to 200° C.

A layer of wood at the surface becomes dehydrated and evolves water vapor with perhaps traces of carbon dioxide,formic and acetic acids (58, 65, 98 ), and glyoxal (33,116 ).

Zone B, 200° to 280° C.

Zone A moves farther into the piece of wood and is succeeded by zone B in which pyrolysis remains slow (35 , 43 , 138). Water vapor, carbon dioxide, formic and acetic acids, glyoxal, and perhaps a little carbon monoxide are evolved (79 , 83 , 95) and additional vapors from zone A pass through. Thus far the reactions are endothermic (58 , 138) and the gaseous products are largely noncombustible. The wood slowly becomes charred (76 , 81 ).

Zone C, 280° to 500° C.

Zones A and B move inward to be succeeded by zone C in which active pyrol- ysis begins suddenly and exothermically (43 , 44 , 58 , 98 , 118 ). The tempera- ture mounts rapidly unless the heat envolved is dissipated.Combustible gases and vapors --notably carbon monoxide, methane, formaldehyde, formic and acetic acids,methanol, and later hydrogen --diluted with carbon dioxide and water vapor (9 , 17 , 59 , 70 , 79 , 95) are enciosed forciblv enough to carry with them droplets of highly inflammable tars (44) that appear as smoke (24). The resi- due left in zone C becomes charcoal. The primary pyrolysis products under- go further pyrolysis and reactions with one another before they escape (59 ,

Rept. No. 2136 -2- 65 , 79 , 118). Such secondary reactions may be catalyzed by the charcoal, which is epecially active when formed at these temperatures (5). The second- ary pyrolysis of the tars is especially strongly exothermic (59). In commer- cial wood distillation, the temperature is kept within zone C until zones A and B vanish after reaching the center of the piece (35), and until the tars are ex- pelled. At this stage the process again becomes endothermic (58 , 118). Smok- ing ceases before 400° C. is reached at the center of the Piece (24). Carboni- zation is considered complete at 400° (58) to 600° C. (35 , 87). The fibrous structure of wood is retained despite serious changes in composition up to 300° C. (13 , 35 , 130). Above 400° C. the crystalline structure of graphite is developed (9 , 19 , 39).

Zone D, Above 500° C.

If the surface temperature continues to rise before carbonization becomes complete, zone D --composed of charcoal-- becomes the seat of still more vigorous secondary reactions in which the gaseous products and tars rising from the zones underneath are further pyrolyzed to more highly combustible products. For example,carbon dioxide and water vapor react with carbon to form carbon monoxide, hydrogen, and formaldehyde (79).

The Course of Combustion

The course of events when wood is heated in air is similarly zonalized but is modified by oxidation reactions and, after ignition, by combustion of pyrolysis and oxidation products.

Zone A, up to 200° C.

The gases evolved by very slow pyrolysis are not ignitible. The wood loses weight steadily if only slowly (76 , 120). Above temperatures as low as 95° C. the wood eventually becomes charred (76 , 81). Oxidation reactions occur that are exothermic and may lead,under conditons in which the heat is conserved, to self-heating and even to self-ignition (6 , 81). Sound wood, however, does not ignite within zone A.

Zone B, 200° to 280° C.

Although the gases evolved still are not readily ignitible, an exothermic condi- tion is reached at lower temperatures than in pyrolysis out of contact with air.

Rept. No. 2136 -3- The temperature at which the overall reactions of pyrolysis and oxidation be- come detectably exothermic has been taken as one of several definitions of the ignition point of wood (11 , 131), even though spontaneous flaming does not begin until higher temperatures are reached (24). The exothermic point has been reported variously as 235°to 240° C. (131), 232° to 260° C. (37), 192° to 220° C. (11), and even as low as 150° C. (24). The variations indetermina- tions come lagely from the fact that time as well as temperature is involved in ignition at such low temperatures (26 , 46 , 81,131).

Prince (103) and McNaughton (81) reported that the gases evolved from wood can be ignited by a pilot flame 180°C. after heating for 14 minutes or at 250° C. after heating for 4 to 9.5 minutes,but their measurements may be seriously in error. These temperatures were measured in an air space, freely ventilated by convection,between the wood surface and a hot plate, the temperature of which was not recorded but must have been significantly higher. Absorption of radiant energy and heat evolved by oxidation in the exposed wood undoubtedly raised its temperature well above that of the moving air current. The wood charred before it inflamed and the charcoal layer was highly absorp- tive of heat and exothermically reactive with oxygen (5 , 90). The actual tem- perature of the wood surface when flaming began was probably well beyond the limits of zone B.

Fons (26), by measuring the temperature at two different depths beneath the surface of ponderosa pine and applying the equation for thermal conduction, found that the temperature of the wood at the exposed surface was 343° C. when it burst into flame. In air between 232° and 443° C. the wood was re- duced to charcoal without flame, after which the charcoal began to glow, still without flame.

Zone C. 280° to 500° C.

The mixture of gases copiously evolved in zone C at first should be too rich in carbon dioxide and water vapor to sustain flame, but it soon becomes more combustible as a result of secondary pyrolysis (79). The gases may then be readily ignited by a pilot flame and burn steadily in luminous diffusion flames. At this stage flaming combustion occurs entirely in the gas phase outside the wood because the rapidly emerging gases lack necessary oxygen until they have gone far enough to mix with air in proportions between the lower and upper limits of flammability (77). Indeed, under suitable conditions the flaming com- bustion may occur at a considerable distance from the wood.

Self-sustaining diffusion flames from organic fuels burn at 1100° C. or some- what more (30). One-half to two-thirds of the heat of combustion of wood is

Rept. No. 2136 -4- liberated in flaming,the balance in glowing combustion of the charcoal (14). How much the heat of flaming contributes to the spread of combustion depends on such external conditions as the geometry of the situation and the velocity of the air. A vertical stick less than 3/4 inch thick that is ignited at the lower end in still air may burn to the top but, in a strong wind or if ignited at the upper end, most of the heat is dissipated and the fire may go out. If the stick is more than 3/4 inch thick, it probably will not burn to the top even if ignited at the bottom in still air, unless external heat is applied (14). A single log in a fireplace will not keep burning without continuous supply of heat from flam- ing kindling or hot coals beneath it,but three logs properly spaced may keep burning as long as each contributes heat to the others.

As long as gases pour forth rapidly enough to blanket the wood surface to the exclusion of oxygen,the charcoal formed cannot burn and is left to accumulate. Since charcoal has only one-third to one-half of the thermal conductivity of wood, the layer of charcoal retards penetration of heat and delays attainment of the exothermal point in the wood underneath (84 , 113). Thus, after the first vigorous flaming of wood,there is often a diminution of flaming until sufficient heat has passed through the insulating layer of charcoal to pyrolyze deeper portions of the wood.

If there is no pilot flame or its equivalent to ignite the gases, there may be no flaming until pyrolysis in zone C is nearly complete and the emission of gases slows sufficiently to let air make contact with the charcoal layer. Since charcoal does not melt below 3480° C. or boil below 4200° C. (67), it cannot go forth to meet the air for its combustion,but must burn in place as air reaches it. Nevertheless, charcoal is an excellent absorber of radiant heat and has a lower temperature of spontaneous ignition than any of the other major products of pyrolysis. Thus temperatures of spontaneous ignition for mixtures with air between the limits of flammability are reported as 644° to 658° C. for carbon monoxide, 580° to 590° C. for hydrogen, 566° C. for acetic acid, 539° to 750° C. for methane, and 430° C. for formaldehyde (67). Pine tar is said to ignite spontaneously at 355°C. But for charcoal spontaneous ignition is re- ported at temperatures as low as 150° to 250° C. (5 , 26 , 90 , 102). Even at 100° C. the combination of charcoal with oxygen generates 2.7 calories per cubic centimeter of oxygen reacting (5). The hot charcoal from the industrial distillation of wood in retorts takes fire if it is allowed to come in contact with air before it cools (43). When wood ignites spontaneously, it is therefore the charcoal that takes fire first (44 , 26) late in zone C after the first vigor of emission of gases is over and air can reach the solid surface. The best meas- urements of surface temperatures for spontaneous ignition of wood agree with this view --Lawson and Simms, 350° to 450° C. (68); Fire Research Board, 380° C. (24); Fons,343° C. (26 ); Hawley, above 300° C. (44); Jones and Scott, 270° to 290° C. (52).

Rept. No. 2136 -5- Zone D, Above 500° C.

At 500° C. (incipient red heat) the charcoal glows and is comsumed. The in- terior of the piece of wood may still be in zones B or C at much lower tempera- tures (3). When the surface rises somewhat beyond 1000° C. (yellowish-red heat), carbon is consumed at the surface as fast as the reaction zones pene- trate into the piece (79). The luminous diffusion flames give way, as the re- actions of primary wood pyrolysis become exhausted, to the nonluminous dif- fusion flames of burning carbon monoxide and hydrogen. When the supply of carbon monoxide and hydrogen finally fail,the remaining charcoal merely glows with little or no flame.

Pyrolysis Reactions

The relative proportions of gases, vapors, tars,and charcoal and the rela- tive proportions of flammable and nonflammable gases produced will vary widely according to the conditions of temperature, pressure, time, geometry, and environment under which pyrolysis occurs. The yields of products may also be altered greatly by the presence of retardants or combustion catalysts in the wood or on its surface.

In carbonizing birch from an initial temperature of 250° C. to a final tempera- ture of 400° C. at different rates and pressures, Klason (59) found the varia- tions in yields of charcoal, tars, water, carbon dioxide, and carbon monoxide reported in table 1. Here pyrolysis at a pressure of 5 millimeters of mercury revealed very nearly the primary pyrolysis products because they were re- moved from the reaction zone before there had been much time for secondary pyrolysis. Less than 20 percent of the wood was left as charcoal and nearly 40 percent was converted to tars,which are highly combustible. At atmos- pheric pressure the yield of charcoal rose and that of tars declined, the more so the slower the pyrolysis. For very slow pyrolysis the yield of charcoal doubled, whereas that of tars nearly vanished. Secondary pyrolysis therefore converted tars into gases and tar coke, the latter being weighed as part of the charcoal. Careful computation of the thermal balance (56 , 59) revealed that the primary pyrolysis in vacuo is not exothermic but consumes heat. The liber- ation of heat under ordinary conditions comes from strongly exothermic pyrol- ysis of the tars. The secondary reactions increase the quantities of water and carbon dioxide in the gaseous products without increasing the quantity of car- bon monoxide. Thus rapid pyrolysis,such as necessarily occurs when wood burns,produces more flammable tars and gases, less diluted with water and carbon dioxide, than does slow pyrolysis.

Rept. No. 2136 -6- Industrial distillation of wood in retorts in the course of 16 to 24 hours com- monly yields 40 to 50 percent of charcoal (1 , 43 , 57). Carbonization of saw- dust in 25 minutes in a rotary kiln produces 23 to 26 percent charcoal (106 , 108). Carbonization of sawdust in 7 to 8 minutes in a multiple-hearth furnace gave 21 percent charcoal (34). In laboratory tests, small birch blocks, which were carbonized in less than 1 minute in an atmosphere of nitrogen to preclude loss by oxidation, yielded only 12.75 percent charcoal of composition similar to that of industrial charcoal (34). Moreover, chemicals impregnated into wood may alter the yields of tars and charcoal, as is described farther on.

Rapid heating through the range of active pyrolysis tends to produce little char- coal, much tar,and highly flammable gases that are rich in hydrogen, carbon monoxide, and hydrocarbons;slow heating tends to produce much charcoal, little tar, and less flammable gases in which there is much water and carbon dioxide ( 79 ). In slow heating, decomposition proceeds in an orderly manner in which there is stepwise formation of increasingly stable molecules, richer in carbon and converging toward the hexagonal structure of graphitic carbon (9 , 39 , 87). In very rapid heating,macromolecules may be literally torn into volatile fragments with little possibility of orderly arrangement (79).

The chemical theory of flameproofing wood and other cellulosic materials (15 , 20 , 23 , 36 , 48 , 69 , 114 , 115) is based on changing the pyrolysis mechanism from that of pyrolysis to that of slow pyrolysis. If the pyrolysis of the cellulose could be directed to complete dehydration (69) according to the equa- tion:

there would be no flammable gases from the major component of wood until temperatures were high enough for the water-gas reaction to set in, by which time most of the water would escape.

The slow and fast reactions of pyrolysis for cellulose, together with the heats of formation (DH f ) involved,have been represented schematically (79) as fol- lows:

Rept. No. 2136 -7- Thus slow pyrolysis yields charcoal and oxygenated gases and vapors of low flammability and releases energy,whereas fast pyrolysis yields little or no carbon, forms hydrogenated gases and vapors, and consumes energy.

The pyrolysis of wood (39, 10, 120, 139) and other cellulosic materials (64, 78, 79, 91) follows reasonably closely the kinetics of a first-order reaction. It is "diffusion-controlled" rather than rate-controlled," the rate being deter- mined by the rate of energy transfer within the solid rather than by the rate of pyrolysis (79). There are satisfactory kinetic data for slow pyrolysis of wood below the exothermic region and for other cellulosic materials both below and above the exothermic region. Such data are given in table 2. Martin (79), on the basis of the data of Stamm (120) and of Kujirai and Akahira (64) in table 2, saw evidence of an expected alteration in kinetics on passing from slow py- rolysis below the exothermic region to fast pyrolysis above it. He suggested that the kinetics may be formulated as follows:

Rept.No. 2136 -8- in which w is the weight of wood remaining after time, t; kl, Al, and El are respectively the specific rate constant, frequency factor, and activation energy for the slow pyrolysis, and k2, AZ, and E2 are the corresponding parameters for the fast pyrolysis; R is the gas constant, T the absolute temperature, and e the base of Napierian logarithms. Constant12 would begin to predominate 20 at 400° C. if A2 is about 10 per second and E; is about 50 kilocalories per mole. It is therefore interesting to find in table 2 that Madorsky, Hart, and Straus (78) in later work found activation energies of 45 to 50 kilocalories per mole for various cellulosic materials when pyrolyzed at higher temperatures. Discussion of the effect on activation energy of impregnation with added chemi- cals, as shown in table 2, belongs to the section of this report on possible mechanisms of chemical action of retardants.

Pyrolysis of Wood Components

The yield of products when wood has been completely pyrolyzed is about what would be obtained by pyrolyzing separately the proportional amounts of the major wood constituents -- hemicellulose, cellulose, and lignin (23, 35, 70). Breakdown of the components, however,is not entirely simultaneous. The hemicellulose, particularly its pentosans, are said to decompose first) largely between 200° and 260° C.,followed by the cellulose at 240° to 350° C., and finally by the lignin at 280° to 500° C. (43, 63, 65, 83, 89). Some investigators accordingly report three peaks in the exothermic region (63, 93, 95, 118, 124).

Hemicellulose

Hemicellulose evolves more gases,less tar, and about as much aqueous dis- tillate as are formed from cellulose,but differs from cellulose in that hemi- cellulose yields no levoglucosan (50, 92).

Much of the acetic acid formed in pyrolysis of wood is attributed to the hemi- cellulose (54, 61, 88). Scission of a carbon-to-oxygen bond in a pentose might lead to further splitting to acetic acid and perhaps formaldehyde or carbon monoxide and hydrogen:

Rept. No. 2136 -9- Similar splitting of a hexose could produce three molecules of acetic acid.

Pentoses are also known to yield furfural and other furan derivatives (35, 83, 121). This production can occur readily by dehydration, to which carbohy- drates are highly susceptible:

Cellulose

Cellulose evolves water in the first stage of thermal decomposition before any other significant changes are observable. On the other hand, cellulose triacetate evolves acetic acid instead of water and fails to yield water even when pyrolysis is completed (78). Therefore, dehydration occurs with cellu- lose and deacetylation occurs with cellulose triacetate, presumably at random, along the chains of glucosan units.

Early in the pyrolysis of cellulose some of the carbon-to-oxygen bonds in the links between glucosan units may be expected to break at random points along the chain. Even at room temperature, cellulose is easliy hydrolyzed at these points, especially so in the presence of acids. In pyrolysis, water and acids are present from prior pyrolysis of the hemicellulose and from dehydration of the cellulose. Besides hydrolysis, pyrolytic scission may serve to break the links. The points in the cellulose macromolecules at which the first ran- dom breaks in the carbon-to-oxygen bonds take place and the way in which the

Rept. No. 2136 -10- the fracture is brought about may govern very largely the subsequent course of pyrolysis of the fragments so produced.

Hydrolysis at one carbon-to-oxygen link in a cellulose macromolecule produces two shorter macromolecules, the new end groups of which differ in properties. One is a “reducing” end group that can dissociate to an open-form polyhydroxy aldehyde, whereas the other is a “nonreducing” secondary alcoholic end group ( 79 ):

(b) glucosan unit with “re- (c) glucosan unit with ducing” end-group that “nonreducing” end- dissociates to an open- group that cannot chain aldehyde dissociate like (b)

Rept. No. 2136 -11- The interior glucosan units (a), the units with “reducing” end group (b), and the units with “nonreducing“ end group (c), may be expected to behave differ- ently on pyrolysis. Still further differences in behavior would be expected if the glucosan units had lost one or more molecules of water by dehydration be- fore the hydrolytic cleavage of the carbon-to-oxygen links took place.

If both carbon-to-oxygen links of an interior glucosan unit (or the one link of a unit on the end of the chain) is hydrolyzed,a molecule of glucose is produced. Glucose can dehydrate to a -glucosan (1, 2 anhydroglucosan), which above 110° C. rearranges to b -glucosan or levoglucosan (1, 6 anhydroglucosan) (51, 80, 100).

Levoglucosan is a characteristic product of the primary pyrolysis of cellulose and plays an important part in one of the modern theories of flameproofing cellulosic fabrics (115). Yields of levoglucosan as great as 50 to 53 percent, representing about one of every two glucosan units, have been obtained from pyrolysis of cellulose at 200°to 300° C. in vacuo (78, 99, 132), and it has been found in lower yields at atmospheric pressure (78, 116, 126). Presence of water-soluble inorganic impurities reduces or prevents formation of levo- glucosan (132) and it is not formed by pyrolysis of wood unless lignin is first removed, at least in part, by chlorination (92). Levoglucosan melts at 180° C., remains stable up to 270° C.,and then pyrolyzes to water, formic acid, acetic acid, and phenols (100).

It is possible that the initial break in carbon-to-oxygen chain links may occur by random pyrolytic scission (78) in stead of hydrolysis. If the breaks occur in bonds between oxygen and number 1 carbon atoms, levoglucosan may be formed by double Walden inversion (80) with hydrogen atoms from alcohol groups of adjacent glucosan units wandering to the oxygen atoms thus servered:

Rept. No. 2136 -12- The anhydroglucose unit marked (d) resembles the (b) glucosan unit with “re- ducing”end group obtained in hydrolytic splitting of the chain. The (d) unit perhaps might open into polyhydroxy acids or polyhydroxy ketones of some such configurations as:

The glucosan unit marked (c) from pyrolytic scission is identical with the (c) unit from hydrolysis.

If pyrolytic scission breaks the bond between oxygen and number 4 carbon atom,instead of number 1, formation of levoglucosan is not possilbe. Mad- orsky, Hart, and Straus (78) consider that the maximum yield of approximate

Rept. No. 2136 -13- one molecule of levoglucosan for each two glucosan units indicates an even chance of pyrolytic scission on either side of the oxygen link. Scission be- tween oxygen and carbon atom 4 presumably leaves the configuration so un- stable that the glucosan ring decomposes to simple compounds such as water, carbon dioxide, and carbon monoxide,which always appear among the products. The same authors observe that neither the degree of crystallinity nor the de- gree of polymerization of the cellulose seem to affect pyrolysis appreciably because cotton cellulose, hydrocellulose made from it, viscose rayon, and cellulose regenerated from cellulose triacetate all pyrolyze at similar rates, with similar activation energies and similar yields of products.

After the initial breaking of cellulose macromolecules by hydrolysis or by pyrolytic scission,subsequent pyrolysis may proceed differently for the frag- ments that differ in their end groups. Open-chain polyhydroxy aldehydes, ke- tones,or acids formed by dissociation of fragments with “reducing” end groups, marked (b) or (d) in the preceding, might readily pyrolyze to simpler hydroxy aldehydes, ketones, or acids (79). Simultaneous oxidation-reduction of adja- cent aldehyde and alcohol groups to form hydroxy acids or hydroxyketones is possible (50). In such ways it is possible to account for formation of formal- dehyde, acetone, glyoxal, glycolic aldehyde, glycolic acid, lactic acid, dilac- tic acid, formic acid, and acetic acid, as well as water, carbon monoxide, and carbon dioxide, all of which have actually been obtained in substantial quan- tities from pyrolysis of cellulose (35, 116).

On further pyrolysis at higher temperatures,carboxyl groups yield carbon dioxide, a nonflammable flame-inhibitive gas,whereas aldehyde groups yield formaldehyde and then hydrogen and carbon monoxide, a highly flammable mixture.

Rept. No. 2136 -14- The “nonreducing”end group marked (c) produced in hydrolysis or in pyro- lytic scission differs from the interior groups marked (a) in having one ex- tra unassociated alcoholic group (79). In the pyrolysis of units with either (a) or (c) groups, by analogy with other polyhydric alcohols, adjacent alco- holic groups might be expected to eliminate water:

to form a ketone. In fact, dehydration of the cellulose in the very first stage of thermal decomposition may produce such ketones before hydrolysis or py- rolytic scission begins. Ketones commonly decompose thermally to ketenes (50), the simplest of which is formed by thermal decomposition of acetone:

Ketenes have not been reported among the pyrolysis products of cellulose, but they are so reactive that no more than transient existence is to be expected. Nevertheless they may exert much effect on the course of pyrolysis. Since ketenes oxidize rapidly in air even at low temperatures, they may be expected to afford important sources of ignition.

More drastic dehydration of glucosan units may produce furan derivatives much as hydroxymethylfuraldehyde is known to result from loss of three molecules of water from glucose. Furaldehyde and furans have been identified among the pyrolytic products of wood (35, 83, 110, 121). Presence of catalysts that stim- ulate dehydration would be expected to favor such reactions.

Breaking of carbon-to-carbon bonds in glucosan units leaves unstable configur- ations that must rearrange or shed such fragments as CHOH or CHOH · CHOH, which must then take stable forms or break up further. CHOH · CHOH can form acetic acid (CH3COOH), separate into ketene (CH2 CO) and water, or into methane and carbon dioxide.

At the temperatures of pyrolysis,the transitory existence of free radicals that can then participate in chain reactions is highly probable. Among such possible free radicals are CHO, CH 3, CH2 , H and OH. Because of their short lifetimes they could not be found as pyrolysis products, but they may well affect the course of pyrolysis reactions. In particular, the free radicals may exist long enough in the vapors from pyrolyzing wood to play a part in ignition in the region where the evolving vapors. mix with air.

Rept.No. 2136 -15- Not all of the pyrolytic reactions in cellulose proceed in the direction of break- up into smaller molecules and molecular fragments. There is a conver- sion of part of the cellulose into a substance closely resembling lignin (33, 43, 83, 89, 110) and some aromatization and aromatic condensation of the decom- position products (39, 65). This may furnish the clue to formation of high- boiling components of the tars produced in pyrolysis. For example, phenolic compounds presumably would condense readily with aldehydes to form resin- ous substances.

Lignin Lignin on pyrolysis is particularly productive of aromatic products (25, 28, 54, 65 and yields more charcoal than iS obtained from cellulose (8 ,25, 35, 70). Lignin does not form a characteristic major product of its intial pyroly- sis corresponding to the levoglucosan from cellulose. Such Difference might be expected since lignin presumably is built up of a variety of simpler units, whereas cellulose is more nearly a condensation polymer of repeated, identi- cal units. Nevertheless,the primary products of pyrolysis of lignin are closely related structurally to the formulas attributed to lignin (28).

Rept. No. 2136 -16- Products formed from aromatic nuclei:

Products formed from straight- chain fragments:

Apparently initial breakdown occurs chiefly in the straight-chain links con- necting such aromatic units as the vanillyl, syringyl, and guaiacyl groups. The aromatic units give rise to phenols, xylenols, guaiacols, cresols, and catechols. the streaight-chain links produce carbon dioxide, hydrocarbons, formic acid, acetic acid, higher fatty acids, the methanol (8, 28, 35).

The pyrolytic products first formed from hemicellulose, cellulose, and lignin promptly undergo further reactions, not all of which are pyrolytic. Polymeri- zations and condensation reactions to form more complex molecules -- such as high-boiling tars, waxes and resinous substances with perhaps phenol- formaldehyde-type lingkages -- are also involved (25, 35, 39, 65, 118). Goos (35) lists 213 compounds that have been indentified in the liquid products of destructive distillation of wood. -17- Rept. No. 2136 The great complexity of the primary pyrolysis of wood and the succession of secondary reactions readily account for the wide variations in yields of gases, vapors, tars,and charcoal according to the particular conditions under which pyrolysis takes place. Such variability in results, however, offers hope that sufficient knowledge of the factors that control the course of the primary py- rolysis might lead to improved methods of making wood more resistant to fire. In this connection it is especially significant that small amounts of im- purities or additives are already known to exert profound effects. The yield of levoglucosan in pyrolysis of cellulose is very sensitive to water-soluble inorganic impurities in the cellulose (78, 132). As little as 1.3 percent of sulfuric acid in lignin increases the yield charcoal and decreases the yield of tars (25). Treatment of wood with phosphoric acid (41 ,96 ) or with mag- nesium chloride, zinc chloride,aluminum chloride, iron sufide, or cobalt sulfide (7) increases the yield of charcoal and decreases the yield of tars. The first four of these chemicals, at least, have been found to exert fire- retardant action (49, 128, 129). On the other hand, sodium carbonate (41), lime, alumina, thoria, zinc oxide,and chromium oxide either decrease the yield of charcoal (7) or have little effect on the yield (112), and have been found inferior in fire retardance (49) when used as impregnants. But in py- rolysis of cotton cellulose,impregnation with sodium carbonate or with sodium chloride was found to increase the yield of charcoal and decrease the yield of tar (78). Potassium carbonate, however,has proved effective in extinguish- ing fire (128, 129).

Formation of Charcoal

Nearly all of the simpler gaseous compounds that are split off in the earlier stages of pyrolysis,where they escape readily, are richer in hydrogen, in oxygen,or in both, than the original cellulose or lignin. The nonvolatile resi- due therefore becomes enriched in carbon. Above 300° (13) or 400° C. (9, 19), the carbonaceous residue assumes a hexagonal graphitic network in which the carbon-to-carbon bonds are unbreakable by pyrolysis alone to tem- peratures beyond 3000° C. Such primary charcoal perhaps retains many of the carbon-to-carbon bonds originally present in the cellulose and lignin.

The more complex tarry products of pyrolysis, which escape less readily from the solid residue,undergo further strongly exothermic (59) pyrolysis that likewise leaves a more highly carbonaceous residue (39, 65) and adds a secondary charcoal or tar coke (59) to the primary charcoal first formed. The extent to which tars are converted to secondary charcoal presumably depends largely on the length of time during which they remain in the solid residue (59 ).

Rept. No. 2136 -18- The charcoal formed up to the time of completion of active pyrolysis and cessation of rapid production of gases and liquids still retains some hydro- gen and oxygen. On heating to still higher temperatures, it slowly gives off volatile material, chiefly hydrogen and carbon monoxide, and the density in- creases up to 800°or 1000° C. (33, 59, 87, 101). Commercial charcoal com- monly contains about 14 percent volatile matter (34).

Both the available data for pyrolysis of wood and the speculations about the possible pyrolytic reactions of wood components indicate that the relative yields of gases, tars,and charcoal and the proportions of flammable and non- flammable gaseous products are subject to wide variations, and may be direc- ted by suitable catalysts or treatments. In general it is desirable, for fire resistance, to direct pyrolysis toward maximum production of charcoal, mini- mum production of tars,and the highest possible proportions of water, carbon dioxide, and highly oxygenated substances in the volatile products. A speed- ing up of the earlier, relatively slow,and less drastic part of the pyrolysis would work toward these ends, provided that deterioration of the wood did not begin within the range of temperature in which wood is customarily used. Dur- ing the slower,cooler portion of the pyrolysis range, hydrolysis and dehydra- tion reactions can proceed in more orderly fashion to denude the still macro- molecular cellulose and lignin fragments of hydrogen and oxygen and of sub- stituent and side-chain groups. Thus there will be less disruption of carbon- to-carbon bonds in glucosan and aromatic rings, leaving time for the carbon residues to condense into charcoal. The later, fast pyrolysis at higher tem- peratures proceeds much more violently, with more disorderly disruption of ring and chain structures to form tars and more highly hydrogenated gases, and with less condensation of carbon to form charcoal.

Theories of Flameproofing

Herodotus (45) wrote that the ancient Egyptians steeped wood in alum solution to impart resistance to fire. In the days of the Roman Empire, Remans soaked wood in vinegar and alum (32, 48, 135) or coated it with clay, lime, and loam, probably in an organic binder (32, 111, 135). In 1638 Nikolas Sabbattini rec- ommended paint containing clay or gypsum for Italian theaters (111), and in 1820 Fuchs painted wood in the Munich Theater with sodium silicate (27, 113). In 1735 Jonathan Wild received a patent in England for a treatment with alum, ferrous sulfate, and borax (111). Gay-Lussac (31) in 1821, at the request of Louis XVIII of France,tried many treatments for flameproofing cellulosic fabrics and recommended ammonium phosphate,a mixture of ammonium phos- phate and ammonium chloride,or a mixture of ammonium chloride and borax.

Rept. No. 2136 -19- Thus our best chemicals for impregnating wood for fire resistance are more than a century old. Meantime long lists of chemicals, as many as 400 in a single study (125), have been tested empirically (49, 73, 104, 122, 123, 128, 133, 136) without significant improvement as far as treatment of wood is con- cerned, except perhaps for the discovery that boric acid, when added to borax, imparts resistance to afterglow (60).

Since flaming combustion and glowing combustion occur at different times and places and by distinctly differing mechanisms, they may be expected to differ in the means by which they may be controlled. Effective flameproofing agents may fail to retard glow and vice versa. They must therefore be considered separately. Flaming occurs earlier than glowing, progresses more rapidly, and is more important in the spread of fire.

Theories of flameproofing may be classified as coating theories, thermal theories,gas theories,or chemical theories (71). The theories, however, are by no means mutually exclusive because two or more of them may be, and probably are, operative in a given case.

1. Coating Theories

One of two theories advanced in 1821 by Gay- Lussac (31), and still current, (4, 20, 23, 36, 38, 48, 72, 98, 104, 114, 129) holds that escape of volatile combustion products from the wood and access of oxygen to the wood can be prevented by treatment with chemicals that melt and coat the wood fibers with a liquid or glassy layer before the temperature of active pyrolysis is reached. Mixtures of borax and boric acid may act in that way, although either compon- ent alone leaves a discontinuous,crystalline deposit rather than a glaze on the fibers (104).

Melts that form foams that are stable at pyrolytic temperatures (98) may be more effective than glazes. The foam serves as a barrier to air and flame, provides thermal insulation, and entraps volatile tars. Retention of the tars encourages their secondary pyrolysis to char, thereby reducing the volume of volatile combustibles. Perhaps most effective would be a mixture of compon- ents, one of which melts near 200° C. and evolves nonflammable gases such as water,carbon dioxide, ammonia, or sulfur dioxide, and another that melts just before active pyrolysis begins,producing a foam that remains stable up to 500° C. Coppick (20) reports a fairly close parallel between the height of foam produced by 5 grams of certain flame retardants when heated for 5 min- utes at 450° to 500° C. in a l-inch test tube, and the effectiveness of the re- tardants for cotton cloth as measured by the “45° microburner test” (71). The data appear in table 3.

Rept. No. 2136 -20- The coating theory does not account for the action of all flame retardants that are known to be effective because such useful materials as the ammonium phosphates,ammonium sulfamate, and the ammonium halides have negligible foaming tendencies. The published reports, however, do not seem to have considered the possibility that a suitable impregnant of low melting point might become foamed by the gases evolved in the earlier stages of wood py- rolysis. Perhaps Coppick’s test for foaming tendency should be made with a disk of wood under the charge of chemical in the test tube. Further, it ap- parently has escaped notice that the tars formed in the pyrolysis of untreated wood readily make foam that is stable to high temperatures. Klason (59) ob- served that the secondary pyrolysis of the tars formed in the primary stage of pyrolysis produced a large volume of foam, beginning at 275° C. Once formed in the outer zone of the wood, such foam should have many of the same properties of insulation and entrapment of tars, which will be evolved subsequently, as are claimed for foaming fire retardants.

One promising new fire retardant has reportedly been developed in Germany by the I. G. Farben Company. The material can be brushed or sprayed on the surface of wood, sinks into the wood far enough to effectively disappear, but intumesces tremendously when flame strikes, thereby insulating the wood.

Complete impregnation or even very deep penetration of the flame retardant should be unnecessary for operation of the coating theory. A glaze or foam of moderate thickness in the surface layers or on the surface of the wood should suffice, provided that it remains stable and adherent at flame tempera- tures.

Indeed, fire-retardant paints or varnishes are confined almost entirely to the wood surface. Even the ordinary oil paints and varnishes exert some slight retardant action,although lacquers made with highly combustible nitrocellu- lose do not (91). The fire-retardant coatings contain retardant ingredients that may function in accordance with the thermal, gas, or possibly even the chemical theories in addition to their interposition of a barrier between air and wood to prevent access of oxygen and to hamper the escape of combusti- ble gases. Coatings are said to serve in part by conducting heat away from an igniting source (84) and, on the contrary,to insulate the wood against heat (48). As long as the coatings remain intact,varnishes and paints that are not too highly pigmented are nearly impenetrable to gases (12) and, on dry wood, can withstand pressure beneath them in excess of 500 pounds per square inch (66) without parting from the wood. But coatings that remain hard up to the temperature of active pyrolysis of wood are inclined to crack and scale off. The most effective fire-retardant paints are those that intumesce strongly somewhat before the wood underneath reaches the temperature or active py- rolysis, producing a voluminous foam that remains stable and relatively

Rept.No. 2136 -21- noncombustible up to flame temperatures. A commercial paint of that type was recently tested in the tunnel furnace at the Forest Products Laboratory; a coating 7 roils thick intumesced to an adherent layer of carbonaceous foam more than one inch thick. The foam burned exceedingly slowly where directly in contact with the pilot flame and exhibited no afterglow. That the paint func- tioned largely by thermal insulation was evident from the very thin layer of charred wood underneath. The paint was reported to contain an organic phos- phorus compound that left its phosphorus in the char after pyrolysis.

Paints and other superficial treatments are attractive for the ease with which they can be applied, even after woodwork has been erected, and for the com- paratively small amount of protective material required. But if external heat is supplied long enough, pyrolysis of the wood underneath must eventually set in. Deep impregnation of the wood therefore seems to offer greater possibili- ties.

2. Thermal Theories

Thermal theories are of three kinds -- thermal insulation to retard the access of heat to the wood, thermal conductivity to dissipate the heat more rapidly, and thermal absorption to reduce the amount of heat available for pyrolysis.

(a) Thermal Insulation. -- Thermal insulation by coatings, glazes, and foams has already been mentioned as part of the coating theories. Untreated wood in pieces that are not too small insulates itself thermally by forming a layer of charcoal on the surface to retard the penetration of heat (84, 113, 114). Little consideration seems to have been given to the possibility of further im- proving the insulating effect by treatments to cause intumescence of the char- coal layer to increase its thickness and its insulating value; however, Metz (84) reports that an unusually loose layer of charcoal is formed by salts, such as potassium carbonate, cyanide,or acetate, which presumably release strong alkali at pyrolysis temperatures. In some tests at the Forest Products Laboratory, great swelling of the charcoal on wood treated with potassium carbonate has been observed.

(b) Thermal Conductivity. -- At the opposite pole is a theory that combustion may be prevented by increasing the thermal conductivity of wood sufficiently for heat to be dissipated faster than it is supplied by an igniting source (20, 38, 84, 114). The theory obvioulsy is derived from the funtioning of the Davy Safty Lamp. No experimental evidence has been advanced to support the hypothesis. Presumably, wood's low conductivity would have to be raised to a level approaching that of metals to dissipate heat fast enough for the pur- pose.

Rept.No. 2136 -22- A crude test was made at the Forest Products Laboratory with some small specimens of untreated wood and of similar wood completely impregnated with a metal alloy that would melt at 105° C. Both the untreated and the impreg- nated wood were heated at the center of one face with a tiny flame from a Bun- sen burner. The rate of rise of temperature was observed with a thermocouple in contact with the unheated side,immediately opposite the point of application of the Bunsen flame. Rise of temperature was slower over the metal-impreg- nated specimen until the metal melted,after which the rate of rise was essen- tially the same for treated and untreated specimens until the point of exothermic Pyrolysis was reached. The exothermic point occurred at the same temperature and at nearly the same time for both treated and untreated specimens. The un- treated specimen then quickly burst into flame, whereas the metal-impregnated specimen merely began to smoke and then charred without flaming. Although impregnation with metal effectively prevented flaming combustion, it clearly did so by a mechanism other than dissipation of heat by conduction. Peschek (98) reported that metal-impregnated wood evolves no combustible gases on heating.

(c) Thermal Absorption. -- The third of the thermal theories postulates decom- position or change of state of the retardant chemical to absorb enough heat to keep the temperature of the wood from reaching the ignition point (20, 23, 36, 38, 48, 72 ). The best example of the theory perhaps is the poor combustibility of wet wood. Few substances, however,have endothermic heats of transition or of decomposition comparable to the heat of vaporization of water. Even with water a high retention of retardant is required for good effectiveness. The protection is at best temporary because exhaustion of the endothermic evaporation leaves the wood subject to combustion as usual. Study of the thermal behavior of a large number of substances failed to reveal any connec- tion between effectiveness as fire retardants and ability to absorb heat (20).

Borax, a good retardant,contains 10 molecules of water of crystallization, which is 89.5 percent by weight of the anhydrous chemical. On heating, it be- gins to lose the water of crystallization at 80° C. and loses all of it at 200° C. The heat of hydration is 34.1 kilocalories per mole (82). The “melting point” of the remaining salt is given variously as 561° to 878° C. If borax prevents combustion wholly by absorbing heat by its dehydration, enough of the chemi- cal must be present to offset, or nearly offset,the heat liberated by pyrolysis of the wood above the exothermic point. Klason (59) calculated the heat liber- ated as 96.7 kilocalories per kilogram for 8-hour pyrolysis and as 169.2 kilo- calories per kilogram for 14-day pyrolysis of birch at atmospheric pressure. For a cubic foot or about 20 kilograms of birch, the heat liberated would be 1,934 to 3,384 kilocalories. To offset it, from 56.7 to 99.5 moles or 25.2 to 44.3 pounds of borax per cubic foot,calculated as anhydrous salt, would be required. Yet borax is effective as a fire retardant in retentions of 3 to 6 pounds per cubic foot (49). Absorption of heat by dehydration, therefore, can account in no more than a minor part for its effectiveness.

Rept. No. 2136 -23- Schlegel (113) tried to establish a parallel between the effectiveness of 30 fire-retardant salts and their“molecular heat capacity” between room tem- perature and the temperature of combustion (taken as about 800° C. ); or, if the salts dissociate, to establish the sum of the “heat capacities” of the com- ponents that exist at combustion temperature. Since the specific heats of the salts at high temperatures are not known reliably, the considerations remain highly speculative and unconvincing (84). 3–

Absorption of heat should prove most beneficial if it occurs at the exposed surface of the wood where the temperature of exothermal pyrolysis is reached soonest. It is conceivable that fire-retardant paints might act partly in this way. As yet, however,there seem to be no data on the thermal changes that occur in such paints.

Although the thermal theories no doubt play minor parts in fire retardance, they do not seem to offer much prospect of providing sustained protection at moderate retentions.

3. Gas Theories

There are two gas theories --that of dilution of the combustible gases of pyrolysis with noncombustible gas es from the retardant, and that of catalytic inhibition of flaming by free radicals capable of breaking the reaction chains of normal gaseous combustion.

3– Schlegel’s work is of interest for his novel adaptation of the Forest Products Laboratory fire-tube test. The apparatus was scaled down to accommodate test specimens 1.0 by 1.0 by 10 centimeters in size. In place of the weight loss after combustion has ended, significance was attached to the average rate of loss in weight during the time taken for the temperature at the top of the chimney to reach its maximum. The effectiveness, W , of a treat- ment with x gram molecules of a given salt per kilogram of wood was taken as the difference between the rate of loss of untreated wood, yo, and the

rate of loss of the treated wood; that is, w = yo - y. For relatively heavy treatments, such as 1 gram molecule or more per kilogram of wood, a “law of mass action” was disclosed whereby x = aw in which a was a de- terminative constant for the particular salt. From this the "specific ef- fectiveness"of a salt was defined as 1 /log a. For light treatments the “law of mass action”usually failed to apply; in fact, 15 of the salts tested actually increased the rate of loss in weight in very light treatments even though some of them --notably ammonium bromide, lithium chloride, and potassium carbonate --were very effective in heavy treatments. Cop- pick (20) similarly reported accelerated combustion of cotton fabric with very light treatments of some salts that were very effective in heavier treatments. Rept. No. 2136 -24- (a) Dilution with Noncombustible Gas. -- In addition to his coating theory, Gay- Lussac (31) offered a second theory to the effect that noncombustible gases evolved by decomposition of the fire retardant will dilute the combustible gases from the pyrolyzed wood sufficiently to render the mixture nonflammable in air. The theory still finds supporters (4 ,20, 23, 36, 38, 48, 72, 84, 98, 113, 114). In a variation of the theory,the diluting gases are thought to form a blanketing layer to exclude oxygen from contact with the combustible pyrolysis gases until the latter have escaped from the zone of high temperature (20, 79).

The gases usually considered effective diluents are water, carbon dioxide, ammonia, sulfur dioxide, and hydrogen chloride. One or more of them may come from decomposition of highly hydrated salts, from alkali carbonates or bicarbonates, from ammonium halides, phosphates, and sulfates, from chlor- ides of zinc, calcium, and magnesium,or from ammonium sulfamate. The retardant must not decompose at temperatures of ordinary use of wood, but it should yield its diluting gases rapidly at a temperature slightly below that at which active pyrolysis of wood sets in, perhaps a little below 270° C. If, as will appear in the discussion of chemical theories,the retardant lowers the temperature of active pyrolysis of wood,the retardant should yield its gases rapidly at a correspondingly lower temperature.

The gas dilution theory has been criticized adversely (20, 104, 129) on the ground that too little noncombustible gas is liberated by any practicable reten- tion of fire retardant to prevent flaming of the large quantity of combustible gases and vapors formed by pyrolysis of wood. The criticism, however, may not be entirely fair because the gases evolved by pyrolysis of wood are already diluted with nonflammable components. Thus, rough calculation from Klason’s data (58) for 8-hour pyrolysis of dry birch at atmospheric pressure indicates that 1 cubic foot of wood yields 590 cubic feet of gases and vapors at 300° C. ; of this amount, 380 cubic feet is steam,75 cubic feet is carbon dioxide, and only 135 cubic feet,or 22.9 percent by volume,consists of flammable gases and vapors. Martin (79), as already mentioned,comments that such a gaseous mixture should be nonflammable. Mixtures in all proportions of carbon mon- oxide with air are rendered nonflammable by addition of 54 percent of steam or of 52 percent of carbon dioxide (21 ). The flammability of the pyrolytic gases from wood therefore comes largely from the mist of tar carried with them. If the 1 cubic foot of dry birch in Klason’s work was impregnated with 6 pounds of borax, calculated on the anhydrous basis,and if the borax did not other- wise alter the pyrolysis of wood,the 10 molecules of water of crystallization in the borax would add about 225 cubic feet of steam to the gaseous products evolved. This, together with the 380 cubic feet of steam and 75 cubic feet of carbon dioxide from the wood,might well serve to suppress the flammability of the tar mist.

Rept. No. 2136 -25- (b) Chain-Breaking Inhibitors. -- A recent theory relies on the inhibition of flaming of the gaseous products by a suitable gaseous catalyst released at py- rolysis temperatures by decomposition of a fire-retardant chemical. The catalysts usually proposed are the halogens and halogen acids, among which the effectiveness has been found to increase in the order of fluorine, chlorine, bromine, and iodine (29, 36, 109, 115). Thus, addition of 6.2 percent of methyl bromide or 2.0 percent of carbon tetrachloride to carbon monoxide makes a mixture that is nonflammable in air in all proportions, whereas addi- tions of 54 percent of water vapor or 52 percent of carbon dioxide are required to accomplish the same purpose by dilution (21). Combustion in flames is believed to proceed by branching chain reactions among free radicals (119). Diluents such as water vapor absorb heat and reduce the frequency of colli- sions between reactive molecules or radicals, but do not alter the chain re- actions otherwise. Inhibitors such as the halogens break the reaction chains. At flame temperatures the halogens decompose to halogen atoms that combine with radicals essential to the combustion process,after which the new com- bination yields a stable molecule,or at least a less reactive radical, and regenerates the halogen to repeat the process (109). Thus, if such free radi- cals as HCO and OH are active in propagating combustion, the presence of a relatively small amount of bromine may destroy them as follows:

Regeneration of the inhibitor permits it to quench flames in much smaller amounts than are necessary with diluents.

Halogen-containing retardants must be stable at temperatures of ordinary use of wood because the halogens and halogen acids attack wood and impair its strength. But the retardant must decompose at or slightly before active pyrolysis of wood begins in order to supply the inhibitor when it is needed. Of course the retardant itself should be incombustible.

4. Chemical Theories

The preceding discussion of the coating,thermal, and gas theories largely ignored any effect exerted by fire retardants on the course of pyrolysis of the wood itself. The chemical theories, on the other hand, aim to change the py- rolysis of wood in a favorable direction. It has already been pointed out that the yields of gases, tars,and charcoal from untreated wood are profoundly altered by changes in the temperature, pressure, or rate at which pyrolysis takes place. That fire retardants likewise alter wood pyrolysis is only to be expected.

Rept. No. 2136 -26- Effective fire retardants significantly lower the temperature at which rapid pyrolysis of wood,with production of char, sets in (16, 85, 94) and raise the temperaturc at which the pyrolysis becomes exothermic (84, 94, 134). The quantity of charcoal produced is materially increased (7, 14, 16, 48, 105, 129, 134). The increase in charcoal must result in decreased production of tar or of gases (14, 48, 105, 123, 134) or of both (129), but the published ports fail to reveal quantitative data from which product balances can be calculated. Much more work, directed toward the theory of fire-retardant action, has been done with cellulose in the form of filter paper (122, 125) cotton textiles (15, 20, 36, 38, 69, 78, 97, 104, 114, 115) than wood. There should, however,be a close parallel between cellulose and wood (114).

Ramsbottom (104) reports the data of table 4 to show how fire retardants lower the charring temperature of cotton fabrics. But marked lowering of the char- ring temperature does not always indicate a corresponding increase in the ate of decomposition of the cellulose (125). The exothermic point in pyrolysis of wood is said to be raised from 300° 500° C. by treatment with fire re- tardants (134). For the effect of chemicals on the yield of charcoal in com- bustion of cotton fabric, Rams bottom (104) gives the data of table 5. Effective retardants may more than double the yield of charcoal, whereas ineffective chemicals, such as sodium chloride,may decrease it. But when pyrolyzed in vacuo at temperatures below 300° C. ,impregnation of cotton cellulose with sodium chloride has been found to increase the yield of charcoal (78).

The increase in yield of charcoal by fire retardants apparently is accomplished chiefly at the expense of tar production. Coppick (20) reported the data of table 6 on yields of dry tar, aqueous condensate, and dry gas for cotton fabric treated with chemicals of varying effectiveness, as measured by a test for duration of afterflaming (71). There, was a close parallel between reduction in flammability and reduction in tar production. Flammability practically ceased when the tar production fell below about 3 milligrams per square centi- meter of fabric,which was a little less than one-fourth the amount of tar from untreated fabric. Borax-boric acid mixture attained that condition and borax alone nearly attained it when the fabric contained about 5 percent of chemical. Sodium chloride,even at 33.7 percent retention,failed to reduce flaming or to reach such a low level of tar production,although it did reduce tar forma- tions substantially. Table 7 from Gulledge and Seidel (38) is of interest in this connection because it reports yields of both char and tar. Nearly all ob- servers of either wood or cellulose fire tests agree that fire retardants sig- nificantly diminish the formation of tar;the exception is Yoshimura and co- workers (140), who seem to report increases in tar formation by zinc chloride, calcium chloride, phosphorus pentoxide, and sodium hydroxide.

Table 6 indicates that added chemicals,whether good or poor retardants, in- crease the amounts of aqueous condensate and dry gas formed on combustion

Rept. No. 2136 -27- of fabric, and do not alter appreciably the composition of the dry gas. Am- monium phosphate and ammonium sulfamate affected the production of gas much like the other chemicals (20). Although added chemicals increased the amount of dry gas without appreciable change in ratio of combustible gases to carbon dioxide, it must be remembered that the gas phase arising from the burning fabric included the tars and the aqueous condensate, which was chiefly water. Since 1 milligram of steam occupies more than 1.25 cubic centimeters, the gas phase was in fact greatly diluted with steam and impoverished in the highly flammable tars. Moreover, with effective retardants the gases and vapors begin to come off more rapidly at lower temperatures than with un- treated fabric, and have more opportunity to escape before taking fire (36, 114).

It appears, therefore, that the effective fire retardants act chiefly by direct- ing the decomposition of cellulosic materials, presumably including wood, toward the formation of much less tar and more water and charcoal, and by starting the decomposition at lower temperatures. Moreover, the retardants probably raise the temperature at which the decomposition becomes exother- mic. The outstandingly important criterion of flammability seems to be the production of tars. The quantity of fixed gases and their ratio of combustible to incombustible components seems to be less important.

Possible Mechanisms of Chemical Action of Retardants

The effect of fire retardants on the charring temperature, the yield of products, and the exothermic point suggests that wood is attacked chemically and partly decomposed before combustion can begin. The majority of retardants impair the strength of cotton fabric to some extent even at ordinary or only slightly elevated temperatures (20 ), and some materials that would otherwise be ef- fective retardants cannot be used because they weaken wood or cotton too seriously (104). Desirable retardants should be harmless to wood at the temperatures of ordinary use,but should themselves decompose or other- wise generate the effective reagent at a temperature above that of common use but well below that of combustion for untreated wood.

Hydrophilic Nature of the Best Retardants

The highly effective inorganic fire retardants are all soluble in water. Al- though such volubility facilitates application,it has the disadvantage of loss of chemical by leaching if the treated wood is exposed to the weather. Efforts have been made to precipitate insoluble retardants in wood by double treat- ments (103, 127). The results, however, have been inferior to those obtainable

Rept. No. 2136 -28- with water-soluble substances. Such success as there was may have been due more to soluble byproducts or excess reagent left in the wood than to the insoluble precipitate. In cotton fabric the insoluble precipitates can be formed and the soluble products and reagents removed easily by washing. Extensive tests made in that way (104) revealed no insoluble substances comparable in effectiveness to the better water-soluble salts. Ferric oxide, stannic oxide, lead monoxide, manganese dioxide,and ferric chromate were moderately effective against flaming but usually stimulated glowing. Some of the insolu- ble compounds thought to have been formed to advantage in wood (127), such as magnesium ammonium phosphate, proved completely ineffective.

No one seems to have pointed out that the poor results with insoluble sub- stances may be of fundamental significance. It strongly confirms the domi- nance of chemical over purely physical mechanisms in the theory of fire- retardant action. Effective retardants perhaps must have those hydrophilic properties that lead to volubility in water if they are to direct the pyrolysis of cellulose and wood toward maximum production of charcoal and water, rather than of tars and combustible gases. The consideration draws atten- tion to possible reactions at the hydroxyl groups of cellulose and lignin.

Effectiveness of Strong Acids and Bases

Both strong acids and strong bases attack cellulose. It has been observed repeatedly that salts that may be expected to dissociate to form acids or bases on heating usually lower charring temperatures and impart good fire resist- ance to wood (6, 84, 117) and to cotton fabric (15, 18, 20, 69, 104, 114). Most of the effective retardants are salts,either of strong acids—with weak bases, such as ammonium chloride or zinc chloride,or of strong bases with weak acids, such as borax or potassium carbonate. Salts of strong acids with strong bases,such as sodium chloride, are usually ineffective. Phosphoric acid and the ammonium phosphates are effective, as is sodium hydroxide also, but trisodium phosphate is much less effective and sodium dihydrogen phosphate and disodium hydrogen phosphate are of little value. (Of course phosphoric acid and sodium hydroxide would be impracticable because they attack wood at room temperature.) Aluminum sulfate is a good retardant, whereas sodium sulfate is not. On the other hand, sodium carbonate is a fairly good retardant and sodium tungstate is definitely good, whereas am- monium carbonate and ammonium tungstate are not.

In addition to the strong acids and bases,oxidizing agents such as potassium nitrate, metallic chromates, selenium,potassium permanganate, and potas- sium thiocyanate degrade cellulose and are of some value as fire retardants (20, 104, 114, 125), although for the most part they are impracticable. Oxi- dized cellulose begins to pyrolyze rapidly at relatively low temperatures and leaves much charcoal and very little tar (78). Rept. No. 2136 -29- There is reasonably satisfactory agreement among various investigators about the relative merits of the water-soluble inorganic salts for flameproofing wood or cotton fabric (49, 97, 104, 113, 122, 123).

Dehydrating Mechanism of Acids

Serebrennikov (117) held that strong acids, such as sulfuric and phosphoric, encourage pyrolysis to carbon and water by their reactivity with hydroxyl groups and by the avidity with which they absorb water. Phosphoric acid, for example, may react first with the hydroxyl groups of glucosan units in cellu- lose thus:

Presence of the phosphate groups then prevents pyrolysis by way of polyhy- droxy aldehydes, glycolic aldehyde, and ketones, as previously described for cellulose. Instead, the acid in conjunction with part of the water formed by the initial reaction may be expected to cause hydrolysis at the linkages between glucosan units,exposing two additional hydroxyl groups for reaction with the acid. Finally the HCOPO(OH)2 groups may pyrolyze to regenerate the phos- phoric acid and leave carbon and water. The overall reaction would then be:

The fact that cellulose acetate cannot be fireproofed with the retardants used for flameproofing cotton (114) lends support to the theory. Since the hydroxyl groups of the glucosan units are already bound to acetyl groups, the initial re- action with inorganic acid is precluded. Decomposition does not begin until the temperature is high enough for the normal pyrolysis of untreated cellulose to set in. The pyrolysis then follows essentially the same course as that des- cribed for cellulose except that acetic acid entirely replaces the water formed in pyrolysis of cellulose (78 ). At higher temperatures the acetic acid yields carbon dioxide and methane,the latter being flammable.

Rept. No. 2136 -30- Lewis Acids and Bases

Several authors (15, 36, 114, 115) have pointed out that the theory of catalytic de:hydration acquires more general application when acids and bases are de- fined in accordance with the Lewis electronic theory (74). Compounds that contain an element capable of accepting electrons to fill out its outer shell are Lewis acids, and those compounds with an element having a lone electron pair to share are Lewis bases. In water,the hydrogen ions dissociated from ordi- nary acids form hydronium ions:

But the Lewis acids and bases include compounds that do not dissociate hydro- gen or hydroxyl ions. Thus, boron trichloride as a Lewis acid neutralizes triethyl amine:

and aluminum chloride reacts with phosgene:

When such processes involve carbon atoms of organic compounds, carbonium ions or carbanions are formed.

Acids catalyze dehydration of a glucosan unit in cellulose in some such man- ner as the following (114):

Addition of the proton to the oxygen atom of a hydroxyl group disturbs the elec- tronic system by drawing the pair of electrons between carbon and oxygen closer to the oxygen,making it easy for water to split off and leave a carbon- ium ion: Rept. No. 2136 -31- The carbonium ion is unstable and rearranges electronically to regenerate the proton, which can then repeat the process at another electron-rich point:

Elimination of water takes place by removal of a hydroxyl group from one carbon atom and a hydrogen atom from the adjacent carbon atom, leaving the carbon atoms joined by a double bond. Remembering that acids also catalyze hydrolytic splitting apart of the glucosan units,it is easy to see how repetition or concurrent operation of the process strips off the hydrogen atoms and hy- droxyl groups and perhaps the oxygen in the ring without breaking carbon-to- carbon bonds.

Dehydration through carbonium ions is effected similarly by such Lewis acids as aluminum chloride,aluminum oxide, phosphoric oxides, and stannic oxide. Thus (114):

Rept. No. 2136 -32- These considerations elaborate Serebrennikov’s picture of the catalysis of pyrolytic dehydration of cellulose by acids and extend it to known fire retard- ants that are not acids in the ordinary sense. Flameproofing by bases such as sodium hydroxide or by Lewis bases can be explained by a similar mechan- ism involving carbanions (114):

Rept. No. 2136 -33- Hydrozen Bonding Action

Another approach to understanding the mechanism by which fire retardants in. crease the yield of charcoal is attributed to W. A. Sisson by Coppick (20):

“The majority of the effective flame-retarding chemicals contain group- ings that are active in hydrogen-bridging processes. The well-ordered and relatively inactive portion of cellulose is considered to involve the hydrogen bonding between adjacent hydroxyls, and the disordered and active portions are characterized by the bridging of adjacent chains via water molecules. The structural stability of the system is critically related to these intermolecular forces and it is proposed that on con- tact with high temperature sources the bridging media are lost owing to the thermal activity of water at these temperatures, but that the linkages may be maintained if a nonvolatile component of sufficient hydrogen-bonding activity is present. The electronic configurations that promote strong linkages of this type in the cellulose system are These groupings are particularly prevalent in the more efficient retardants such as the phosphates, sulfates, and sulfamates,and the theory proposes that thermal stability and the con- fining of fragments to the solid phase is due to a great extent to the hydrogen-bonding capacity of these compounds. It is further pointed out that the main chemical characteristics of the strong dehydrators and the hydrated salts effective in retardant action is their active hydrogen-bonding power,and although they prefer to stabilize their electronic configurations by bonding with water, in the absence of the latter the cellulosic hydroxyl serves the same function.“

The hydrogen-bonding theory is quoted by other recent workers (36, 72), but is rejected by Schuyten, Weaver,and Reid (114) on the ground that even the strongest hydrogen bonds have a bond energy no more than 9 to 10 kilocalories per mole and are unlikely to exist at temperatures of 400° to 500° C. Coppick (20), however,suggests that the di-and tri-functionality of the phosphates, sulfates, and sulfamates,also provide for cross linking through primary bonds to bind fragments together in the solid phase and hamper splitting off of vola- tile chain fragments. Moreover,such three-dimensional effect may proceed also by polymerization or condensation of aldehydic groupings catalyzed by the strong acid or alkali thermal dissociation products of the retardant. Simi- larly Gulledge and Seidel (38) explain the effectiveness of a titanium oxide and antimony oxide complex on the basis of reactivity with organic hydroxyl groups both by primary valence and chelation.

Rept. No. 2136 -34- Application to Lignin

There seem to be no published theories of the effect of fire retardants on the chemical mechanisms of the pyrolysis of the lignin component of wood. Since cellulose normally pyrolyzes at a lower temperature than lignin and perhaps the lag may be even greater in treated wood,it may be that the lignin enjoys a measure of protection by the layer of charcoal left by the cellulose. The structure of wood,which is retained for some time after pyrolysis sets in (35, 83, 130), accords with such a relation because the cavities through which gases must escape are lined with cellulose-rich walls, with the lignin-rich middle lamella sandwiched between them. Presence of the cellulose char may trap any tars from the lignin and favor their secondary pyrolysis. Freshly formed carbon may well be expected to exert profound catalytic effects on sub- sequent pyrolysis of lignin. Lignin normally forms much more char but per- haps a little less tar,and hence may contribute less to flammability than cellu- lose, even without fire-retardant treatment.

The more or less orderly action of Lewis acids and bases at hydroxyl groups leading to dehydration to carbon and water, as postulated for cellulose, can- not apply without modification to lignin with its irregularly varied structure. In place of the 2 to 1 ratio of hydrogen to oxygen in cellulose, the ratio in lig- nin is more than 3 to 1. Nevertheless lignin presents numerous alcoholic hy- droxyl groups for reactions similar to those attributed to cellulose. The ana- logous reaction at phenolic groups would form the carbonium ion

This could hardly reach a stable configuration without breaking the aromatic ring and shifting hydrogen, for example, to

which would still be unstable and break further into carbon and less unsaturated hydrocarbons. The methoxy groups of lignin are believed to split off as methyl alcohol in normal pyrolysis and presumably would do so even more readily in the presence of Lewis acids. Presence of Lewis acids almost certainly would hamper the phenol-aldehyde reaction and similar condensation reactions among primary pyrolysis products to which formation of tar is largely attributed. On

Rept. No. 2136 -35- the whole it is reasonable to suppose that fire retardants lower the tempera- ture of pyrolysis of lignin,increase the production of carbon and water, and diminish the formation of tars much as they do with cellulose. For the ex. cess hydrogen in lignin beyond two atoms of hydrogen for one of oxygen, it is to be remembered that charcoal is far from pure carbon. Charcoal normally retains at least 14 percent of material volatile at much higher temperatures (34), more than 90 percent of which is hydrogen (42).

Theories of Glow Prevention

The process of combustion of charcoal,the solid phase in the pyrolysis prod- ucts of wood, is essentially independent of the process of flaming combustion of the gaseous and liquid phases. Such independence was well demonstrated by Fons (26), who exposed slender dowels of ponderosa pine in an electric fur- nace at different temperatures. At temperatures somewhat below 232° C. the wood was reduced to charcoal without either flame or glow; between 232° and 443° to 463° C. (according to diameter of specimen) the wood charred and was then consumed by glow without flame;above 463° C. the wood flamed until evolution of volatile combustibles ceased, and then glowed until the charcoal was consumed.

It therefore is not surprising that the effect of any one chemical on flaming may differ markedly from its effect on glowing. In tests of many chemicals at the Forest Products Laboratory (49), only the ammonium phosphates and phosphoric acid proved highly effective against both flaming and glowing. So- dium molybdate,which was highly effective against flaming, and chromic acid and the chlorides of chromium, manganese, cobalt and copper, which were only a little less effective against flaming,all made wood glow more intensely than untreated wood. Such excellent flame inhibitors as borax and sodium ar- senate and such good inhibitors as aluminum chloride and ammonium sulfate, as well as poor inhibitors such as sodium chloride, had no noticeable effect on glowing. On the other hand boric acid, ammonium borate, and sodium am- monium phosphate were excellent glow inhibitors but were poor or only moder- ate inhibitors of flame. Similar findings are reported for cotton fabric (20, 69, 72, 104).

Effective retardants for glow are much less numerous than the effective re- tardants for flame. In wood the good inhibitors of glow are chiefly the ammoni- urn phosphates, phosphoric acid,boric acid, or substances that yield phosphoric or boric acid at pyrolysis temperatures. It is of interest that the glowing of carbon deposits in gasoline engines is prevented by adding certain organic phos- phorus compounds to the fuel. These compounds, such as trimethyl phosphate,

Rept. No. 2136 -36- tritolyl phospfate and ditolyl phosphonate (47), yield phosphoric acid on burn- ing. For some other compounds the reports are conflicting. Thus zinc chlor- ide and chromium chloride are said to prevent glow in cotton yarn (97), but in wood zinc chloride was rated only slightly effective and chromium chloride was found to increase glow (49 ). Silica is reported as effective (38 ) and as in- effective (104) against glow in cotton fabric.

There are both physical and chemical theories of glow retardance.

1. Physical Theories

Exclusion of oxygen from contact with the surface of hot charcoal prevents glow. Intumescent paints, provided that they remain intact at combustion temperatures, accomplish such exclusion of oxygen. One of the best intumes- cent paints tested at the Forest Products Laboratory, however, makes a highly carbonaceous foam; so,the question of glow prevention is merely transferred from the wood char to the paint char. The paint in question was reported to contain an organic phosphorus compound that left the phosphorus permanently in the char.

The coating theory has been advanced for some impregnated glow retardants, such as borax-boric acid mixtures,but it has been shown that there is no con- sistent relation between the proportions of such mixtures that form continuous coatings on char and those that are effective against glow (20, 104). Some of the best glow retardants,such as the ammonium phosphates, are not believed to form coatings.

Theories of glow prevention by the cooling effects of increased thermal con- ductivity or of absorption of heat by endothermal changes in the retardant are subject to the same limitations discussed previously in connection with flame- proofing.

2. Chemical Theories

Glowing combustion of charcoal is similar to the combustion of coke or other pyrolyzed forms of carbon. It involves oxidation at or near the surface at temperatures of 600° to 700° C. or higher. Some (2, 72) believe that the re- action occurs in two stages:

kilocalories per mole

kilocalories per mole

Oxidation of carbon all the way to carbon dioxide evolves 94.39 kilocalories per mole. The primary reaction takes place at the interface between carbon

Rept.No. 2136 -37- and air;the secondary reaction occurs in the gas phase and evolves nearly 2-1/2 times as much heat as the primary reaction. There are other theories that postulate initial formation of carbon dioxide followed by its reduction to carbon monoxide (40), simultaneous formation of the monoxide and dioxide (53), or initial pentration of oxygen into the carbon structure to form graph- itic oxides that then decompose to carbon dioxide and carbon monoxide (86, 107). Detailed reviews of the theories have been published (55 ,107). What- ever the theory, however,the net heat liberated by the combution is deter- mined by the ratio of CO to CO 2 in the products. The evidence indicates that

effective glow retardants greatly increase the ratio of CO:CO 2 and thereby minimize the heat available for propagation of the oxidation. If the reaction can be directed very largely to the monoxide, the heat liberated is only 28 percent of that set free in complete conversion to the dioxide and may prove insufficient to propagate combustion.

The enormous extent to which the CO:CO2 ratio can be altered by catalysts is illustrated by some tests by Arthur and Bowring (2). Wood charcoal and other forms of carbon were burned at 850° C. in a stream of dry air to which the catalysts were added in varying quantities. The conditions were such that without catalyst the carbon burned entirely to the dioxide, with practically no

monoxide. But with as little as 0.2 percent of catalyst, the CO:CO 2 ratios were altered as shown: Molar ratio of CO to CO 2 Catalyst in combustion products

Phosphorus oxychloride 8.4 Phosphorus trichloride 6.4 Chlorine 2.9 Carbon tetrachloride 2.6 Chloroform 2.3 Stannic chloride 2.1 Dichloromethane 1.1 Hydrogen chloride 1.16 Sulfur trioxide 0.52 Iodine 0.59

Increasing the catalyst beyond 0.2 percent produced very little further change in the results. As little as 0.2 percent of water vapor in the air stream, how- ever, largely destroyed the effect of such hydrophobic catalysts as carbon tetrachloride. The poisoning effect of water may be due to the water-gas re- action, Catalysts with a strong affinity for water, such as the phosphorus compounds, were better able to compete with carbon monoxide for the water molecules and thereby retain their catalytic power.

Rept. No. 2136 -38- The catalysts studied by Arthur and Bowring, of course, are too volatile to be directly applicable to the glowproofing of wood, but the experiments sug- gest the mechanism by which more practicable glow retardants act in wood or other cellulosic materials.

Little (72) reported studies of glowing combustion independently of flaming combustion of untreated and treated cotton by first pyrolyzing the specimens in nitrogen at 350°C. and then igniting them in air with a pilot flame, noting the loss in weight during pyrolysis, loss in weight by glow, and time of glow. The results appear in table 8. Ammonium sulfamate and borax, which retard flaming well, greatly increased the extent and time of glowing, perhaps be- cause they increased the amount of char. Boric acid, which did not prevent flaming despite the high production of char,almost completely prevented glow- ing. Monoammonium phosphate prevented both flaming and glowing and was highly effective against glowing at a retention as low as 0.5 percent, although a considerably higher retention was necessary for good flameproofing.

The relative performance of borax and boric acid indicates that the acid or its anhydride must be released at combustion temperatures to prevent glowing. Similarly, among the phosphates the acid is the effective agent as is shown in table 9 (72). To the extent that the acid is neutralized in sodium salts, which are stable at combustion temperatures,it is rendered ineffective against glow- ing.

That the effective phosphates,and presumably other glow retardants, act like the catalysts in the experiments of Arthur and Bowring to increase the CO: CO2 ratio in the combustion products is indicated by the data of table 10 (72 ). Untreated cotton fabric and fabric treated with 5 percent of diammonium hy- drogen phosphate was first pyrolyzed in nitrogen for varying lengths of time; it was then burned in air in a“combustion train” at 500° C. and the content of carbon dioxide and carbon monoxide determined in the emergent gases. From untreated fabric the quantities of CO and CO 2 were about the same for fabric that was partly or completely prepyrolyzed, and the ratio of CO to CO2 was less than 0.2. From treated fabric the production of CO 2 diminished and that of CO increased with length of prepyrolysis until -- for fully prepyrolyzed fabric --the ratio CO : CO 2 attained was nearly 1.5, or 8 to 9 times the ratio for untreated fabric. The effect of the phosphate on the CO: CO 2 ratio was less for partly than for completely prepyrolyzed fabric; presumably because water and hydrocarbons were still formed in the combustion of incompletely prepyrolyzed char to poison the phosphate catalyst. Thus glow retardants at- tain their full effect after pyrolysis has been completed, when glow retardance becomes most useful.

As little as 1 percent of diammonium hydrogen phosphate suffices to alter the CO:CO2 ratio, as indicated in table 10, and heavier treatments cause little

Rept. No. 2136 -39- or no further change (72). Boric acid and other effective glow retardants similarly alter the CO:CO2 ratio,whereas ineffective glow retardants do not.

There is good evidence,therefore, to support the theory that glow retardance depends on catalysis of the oxidation of carbon in which the less exothermic primary reaction is favored at the expense of the much more exothermic secondary oxidation to carbon dioxide. On the other hand the stimulation of glow by chromates, molybdates, halides of chromium, man- ganese, cobalt, and copper (49) and ferric and stannic oxides (69, 104) may be attributed to catalytic promotion of the secondary oxidation of the monoxide or of a primary oxidation of carbon directly to the dioxide. It is reported that many metallic oxides lower the ignition temperature and hasten the burning of carbon (62). It is also reported that manganese dioxide catalyzes the oxi- dation of carbon monoxide to carbon dioxide (137). Moreover the catalytic effect of the metallic oxides in directing oxidation toward carbon dioxide can be poisoned by alkalies even in mere traces (69, 137).

There are at least three possible mechanisms for the action of catalysts in directing the oxidation of carbon and thereby the tendency to propagate glow- ing:

1. The catalysts may alter the energy barriers (heats of activation) of one or more of the oxidation reactions (20):

In such alteration, the glow retardants lower the energy barrier for reaction (a) or raise the barriers for reactions (b), (c), and (d). Glow stimulators raise the barrier for reaction (a) or lower that for one or more of the other reactions.

2. The catalysts may effect regenerative cycles of such reactions as the fol- lowing (attributed to W. A. Sisson by Coppick (20)):

Rept. No. 2136 -40- This cycle would be for a glow-inhibiting catalyst. By analogy, perhaps the following might be for a glow-stimulating catalyst:

3. The catalysts may be adsorbed at active centers on the charcoal to alter its reactivity (2):

The third theory seems to provide for glow retardance but not for stimulation or for alteration of the CO:CO2 ratio when oxidation does proceed. It might, however,be the initial step in the cycles of reactions of the second theory. Thus:

Then

On the other hand, for the glow-stimulating stannic oxide the reaction scheme may be:

Rept. No. 2136 -41- Then

As already pointed out, the glow-inhibiting catalysts that produce a high CO: CO2 ratio with low liberation of heat may become inoperative in the presence of water vapor. The interference may occur through the water gas reaction:

This could be followed by oxidation of the hydrogen to regenerate water, thus setting up a branching chain reaction (72). Pyrolysis and flaming combustion release much water, the quantity of which is further increased if flame “re- tardants are present. During that stage of burning, glow inhibitors have little or no effect. Those retardants like diamrnonium hydrogen phosphate, which inhibit both flame and glow,act by entirely different mechanisms in the two regions of burning. The mechanism of glow inibition comes into play when charring has reached a point at which most of the water has been given off. Successful glow inhibitors must remain with the char at this point, and must have enough affinity for watertO compete with carbon monoxide for it and then remove the water at the beginning of the reaction chain. Since few substances combine these two properties, good glow inhibitors are scarced.

Ammonium sulfate, ammonium, borate, and the ammonium phosphates -- in that order-- are increasingly effective against glow. The ammonium salts de- compose to the corresponding acids or acid anhydrides before temperatures are reached at which wood pyrolyzes rapidly. The acids and thier anhydrides have strong affinity for water and catalyze the dehydration of cellulose to in- hibit flaming by mechanisms described previously. By the time complete char- ring is attained, sulfuric acid is largely orentirely dissipated because its an- hydride is volatile. Little (72), without quoting authority, says that such de- composition of sulfuric acid into sulfur trioxide and water begins at 150° to

Rept. No. 2136 -42- 180° C.;others (67, 82) report temperatures as high as 340° C. In any case the loss of retardant should be serious before conditions favorable for glow- ing combustion of the char are reached. The anhydrides of boric acid and of phosphoric acid, however,are volatile to no more than a limited extent, even at temperatures well beyond red heat. The effective glow retardants are known to be held tenaciously by charcoal (20).

Little (72) suggests the following transitions of boric acid to its anhydride:

Boric oxide begins to soften at 577° C.,becomes liquid at 1300° C., and boils at 1500° C. (82). Boric acid, however,is appreciably volatile with steam even at 100° C. (82). Accordingly, the slight inferiority of the berates to the phosphates as glow inhibitors may come from such loss of boric acid with the steam evolved during the charring stage of decomposition.

For phosphoric acid,Little (72) suggests the following transitions:

Metaphosphoric acid does not vaporize until the temperature exceeds 850° C., after which it vaporizes without decomposition. Thus metaphoric acid remains firmly with the residue after charring,where it can exercise its ability to re- tard glow.

Conclusion

Substantial progress has been made in reaching an understanding of the funda- mental mechanisms of the pyrolysis and combustion of wood and of their al- teration by added chemicals. No single theory completely describes the man- ner in which resistance to fire is imparted to wood. In particular, the mech- anism of protection against flaming combustion differs essentially from that against glowing combustion. For resistance to both flame and glow the older physical theories no doubt remain applicable to a certain extent, at least in some cases,but the newer chemical theories seem to be of greater and more

Rept. No. 2136 -43- general significance. The chemical theories especially offer promise of achieving good resistance to fire with smaller expenditures of chemicals than have generally been found necessary in the past.

All of the theories, however,still remain too highly speculative because there has been insufficient experimental work to establish properly the extent of their validity and the quantitative details of their applicability. But enough has been learned to afford a good foundation for a thoroughly scientific approach to the problem of imparting better fire resistance to wood. Since empirical efforts for more than a century have added so little to our stock of practicable fire retardants for wood, it is high time that the objective be sought from the point of view of fundamental principles.

Rept. No. 2136 -44- Literature Cited

Rept. No. 2136 -45- Rept. No. 2136 -46- Rept. No. 2136 -47- Rept. No. 2136 -48- Rept. No. 2136 -49- Rept. No. 2136 -50- Rept. No. 2136 -51- Rept. No. 2136 -52- Rept. No. 2136 -53- Rept. No. 2136 -54- Rept. No. 2136 -55- Rept. No. 2136 -56- Rept. No. 2136 -57- Rept. No. 2136 -58- 1.-71 Rept. No. 2136 -59- Table l.--Variation in yields of products in carbonization of birch at different rates and under different pressures

Rept. No. 2136 Table 2 .-- Activation energies and frequency factors for weight during pyrolysis of wood, wood components, and other cellulosic materials

Rept. No. 2136 Table 3. --Relation between foaming tendency and flame retardance for cotton fabric

Rept. No. 2136 Table 4 .--Lowering of the charring temperature of cotton fabric by fire-retardant chemicals

Rept. No. 2136 . .

Rept. NO. 2136 Table 6. --Yields of tar, aqueoua condensate, and gases in combustion of treated and untreated cotton fabric

1- With effective fire-retardant treatment there is little or no afterflaming. 2- Sodium chloride and boric acid are not flame retardants. Table 7. --Yields of char, tar, water, and carbon dioxide in the com- bustion of treated and untreated cotton fabric

Rept. NO . 2136 Table 8. --Retardation of glowing combustion of the char from fabric that was first pyrolyzed in nitrogen

Rept. No. 2136 Table 9 . --Relative glow retardance of phosphates

Rept. No. 2136 Table 10. --Effect of diammonium hydrogen phosphate on the gases pro- duced by combustion of char from cotton fabric that had been previously pyrlyzed for varying lengths of time

Rept. No. 2136

Chapte r17 Fire Safety Robert H. White and Mark A. Dietenberger

Contents ire safety is an important concern in all types of F construction. The high level of national concern for fire safety is reflected in limitations and design Fire Safety Design and Evaluation 17–1 requirements in building codes. These code requirements are Types of Construction 17–2 discussed in the context of fire safety design and evaluation in the initial section of this chapter. Since basic data on fire Fire Growth Within Compartment 17–2 behavior of wood products are needed to evaluate fire safety Flashover 17–4 for wood construction, the second major section of this chapter covers fire performance characteristics of wood Containment to Compartment of Origin 17–4 products. The chapter concludes with a discussion of Fire Safety Engineering 17–6 flame-retardant treatments that can be used to reduce the combustibility of wood. Fire Performance Characteristics of Wood 17–6 Ignition 17–6 Fire Safety Design Heat Release 17–7 and Evaluation Flame Spread 17–8 Fire safety involves prevention, containment, detection, and evacuation. Fire prevention basically means preventing the Smoke and Toxic Gases 17–9 ignition of combustible materials by controlling either the Charring and Fire Resistance 17–10 source of heat or the combustible materials. This involves proper design, installation or construction, and maintenance of the building and its contents. Proper fire safety measures Flame-Retardant Treatments 17–12 depend upon the occupancy or processes taking place in the Fire-Retardant-Treated Wood 17–12 building. Design deficiencies are often responsible for spread of heat and smoke in a fire. Spread of a fire can be prevented Flame-Retardant Pressure Treatments 17–13 with design methods that limit fire growth and spread within Flame-Retardant Coatings 17–13 a compartment and with methods that contain fire to the compartment of origin. Egress, or the ability to escape from a References 17–13 fire, often is a critical factor in life safety. Early detection is essential for ensuring adequate time for egress. Statutory requirements pertaining to fire safety are specified in the building codes or fire codes. These requirements fall into two broad categories: material requirements and build- ing requirements. Material requirements include such things as combustibility, flame spread, and fire endurance. Building requirements include area and height limitations, firestops and draftstops, doors and other exits, automatic sprinklers, and fire detectors. Adherence to codes will result in improved fire safety. Code officials should be consulted early in the design of a building

17–1 because the codes offer alternatives. For example, floor areas In protected light-frame construction, most of the structural can be increased if automatic sprinkler systems are added. elements have a 1-hour fire resistance rating. There are no Code officials have the option to approve alternative materi- general requirements for fire resistance for buildings of unpro- als and methods of construction and to modify provisions of tected light-frame construction. the codes when equivalent fire protection and structural integrity is documented. Based on their performance in the American Society for Testing and Materials (ASTM) E136 test, both untreated Most building codes in the United States are based on model and fire-retardant-treated wood are combustible materials. building codes produced by the three building code organiza- However, the building codes permit substitution of fire- tions (Building Officials and Code Administrators Interna- retardant-treated wood for noncombustible materials in some tional, Inc.; International Conference of Building Officials; specific applications otherwise limited to noncombustible and the Southern Building Code Congress International, materials. Inc.). These three organizations are developing a single international building code that will replace the existing In addition to the type of construction, the height and area three model building codes. In addition to the building limitations also depend on the use or occupancy of a struc- codes and the fire codes, the National Fire Protection Asso- ture. Fire safety is improved by automatic sprinklers, ciation’s Life Safety Code provides guidelines for life safety property line setbacks, or more fire-resistant construction. from fire in buildings and structures. As with the model Building codes recognize the improved fire safety resulting building codes, provisions of the life safety code are statutory from application of these factors by increasing the allowable requirements when adopted by local or State authorities. areas and heights beyond that designated for a particular type of construction and occupancy. Thus, proper site planning In the following sections, various aspects of the building and building design may result in a desired building area code provisions pertaining to fire safety of building materials classification being achieved with wood construction. are discussed under the broad categories of (a) types of con- struction, (b) fire growth within compartment, and (c) con- Fire Growth Within Compartment tainment to compartment of origin. These are largely requirements for materials. Information on prevention and A second major set of provisions in the building codes are building requirements not related to materials (for example, those that regulate the exposed interior surface of walls and detection) can be found in publications such as those listed at ceilings (that is, the interior finish). Codes typically exclude the end of this chapter. Central aspects of the fire safety pro- trim and incidental finish, as well as decorations and furnish- visions of the building codes are the classification of build- ings that are not affixed to the structure, from the more rigid ings by types of construction and the use or occupancy. requirements for walls and ceilings. For regulatory purposes, interior finish materials are classified according to their flame Types of Construction spread index. Thus, flame spread is one of the most tested fire performance properties of a material. Numerous flame Based on classifications of building type and occupancy, the spread tests are used, but the one cited by building codes is codes set limits on the areas and heights of buildings. Major ASTM E84, the “25-ft tunnel” test. In this test method, the building codes generally recognize five classifications of 508-mm-wide, 7.32-m-long specimen completes the top of construction based on types of materials and required fire the tunnel furnace. Flames from a burner at one end of the resistance ratings. The two classifications known as fire- tunnel provide the fire exposure, which includes forced draft resistant construction (Type I) and noncombustible construc- conditions. The furnace operator records the flame front tion (Type II) basically restrict the construction to noncom- position as a function of time and the time of maximum bustible materials. Wood is permitted to be used more flame front travel during a 10-min period. The standard liberally in the other three classifications, which are ordinary prescribes a formula to convert these data to a flame spread (Type III), heavy timber (Type IV), and light-frame (Type index (FSI), which is a measure of the overall rate of flame V). Heavy timber construction has wood columns, beams, spreading in the direction of air flow. In the codes, the floors, and roofs of certain minimum dimensions. Ordinary classes for flame spread index are I (FSI of 0 to 25), II (FSI of construction has smaller wood members used for walls, 26 to 75), and III (FSI of 76 to 200). Some codes use A, B, floors, and roofs including wood studs, wood joists, wood and C instead of I, II, and III. Generally, codes specify FSI trusses, and wood I-joists. In both heavy timber and ordinary for interior finish based on building occupancy, location construction, the exterior walls must be of noncombustible within the building, and availability of automatic sprinkler materials. In light-frame construction, the walls, floors, and protection. The more restrictive classes, Classes I and II, are roofs may be of any dimension lumber and the exterior walls generally prescribed for stairways and corridors that provide may be of combustible materials. Type II, III, and IV con- access to exits. In general, the more flammable classification structions are further subdivided based on fire-resistance (Class III) is permitted for the interior finish of other areas of requirements. Light-frame construction, or Type V, is subdi- the building that are not considered exit ways or where the vided into two parts, protected (1-hour) and unprotected. area in question is protected by automatic sprinklers. In other areas, there are no flammability restrictions on the interior finish and unclassified materials (that is, more than 200 FSI) can be used.

17–2 Table 17–1. ASTM E84 flame spread indexes for 19-mm-thick solid lumber of various wood species as reported in the literature Flame spread Smoke developed Speciesa indexb indexb Sourcec Softwoods Yellow-cedar (Pacific Coast yellow cedar) 78 90 CWC Baldcypress (cypress) 145–150 — UL Douglas-fir 70–100 — UL Fir, Pacific silver 69 58 CWC Hemlock, western (West Coast) 60–75 — UL Pine, eastern white (eastern white, northern white) 85, 120–215d 122, — CWC, UL Pine, lodgepole 93 210 CWC Pine, ponderosa 105–230d —UL Pine, red 142 229 CWC Pine, Southern (southern) 130–195 — UL Pine, western white 75e —UL Redcedar, western 70 213 HPVA Redwood 70 — UL Spruce, eastern (northern, white) 65 — UL, CWC Spruce, Sitka (western, Sitka) 100, 74 —, 74 UL, CWC Hardwoods Birch, yellow 105–110 — UL Cottonwood 115 — UL Maple (maple flooring) 104 — CWC Oak (red, white) 100 100 UL Sweetgum (gum, red) 140–155 — UL Walnut 130–140 — UL Yellow-poplar (poplar) 170–185 — UL aIn cases where the name given in the source did not conform to the official nomenclature of the Forest Service, the probable official nomenclature name is given and the name given by the source is given in parentheses. bData are as reported in the literature (dash where data do not exist). Changes in the ASTM E84 test method have occurred over the years. However, data indicate that the changes have not significantly changed earlier data reported in this table. The change in the calculation procedure has usually resulted in slightly lower flame spread results for untreated wood. Smoke developed index is not known to exceed 450, the limiting value often cited in the building codes. cCWC, Canadian Wood Council (CWC 1996); HPVA, Hardwood Plywood Manufacturers Association (Tests) (now Hardwood Plywood & Veneer Assoc.); UL, Underwriters Laboratories, Inc. (Wood-fire hazard classification. Card Data Service, Serial No. UL 527, 1971). dFootnote of UL: In 18 tests of ponderosa pine, three had values over 200 and the average of all tests is 154. eFootnote of UL: Due to wide variations in the different species of the pine family and local connotations of their popular names, exact identification of the types of pine tested was not possible. The effects of differing climatic and soil conditions on the burning characteristics of given species have not been determined.

The FSI for most domestic wood species is between 90 and Additional FSI for many solid-sawn and panel products are 160 (Table 17–1). Thus, unfinished lumber, 10 mm or provided in the American Forest and Paper Association’s thicker, is generally acceptable for interior finish applications (AF&PA) design for code acceptance (DCA) No. 1, “Flame requiring a Class III rating. Flame-retardant treatments are Spread Performance of Wood Products” (AWC 1999). usually necessary when a Class I or II flame spread index is required for a wood product. A few domestic softwood spe- There are many other test methods for flame spread or flam- cies can meet the Class II flame spread index and only mability. Most are used only for research and development or require flame-retardant treatments to meet a Class I rating. quality control, but some are used in product specifications A few imported species have reported FSIs of less than 25. and regulations of materials in a variety of applications.

17–3 Since the fire exposure is on the underside of a horizontal Flashover specimen in the ASTM E84 test, it is not suitable for mate- rials that melt and drip or are not self-supporting. Code With sufficient heat generation, the initial growth of a fire in provisions pertaining to floors and floor coverings may be a compartment leads to the condition known as flashover. based on another test criterion, the critical radiant flux test The visual criteria for flashover are full involvement of the (ASTM E648, Critical Radiant Flux of Floor-Covering compartment and flames out the door or window. The inten- Systems Using a Radiant Heat Energy Source). The critical sity over time of a fire starting in one room or compartment radiant flux apparatus is also used to test the flammability of of a building depends on the amount and distribution of cellulosic insulation (ASTM E970, Critical Radiant Flux of combustible contents in the room and the amount of Exposed Attic Floor Insulation Using a Radiant Heat Energy ventilation. Source). In the critical radiant flux test, the placement of the radiant panel is such that the radiant heat being imposed on The standard full-scale test for pre-flashover fire growth is the the surface has a gradient in intensity down the length of the room/corner test (International Organization for Standardiza- horizontal specimen. Flames spread from the ignition source tion (ISO) 9705, Fire Tests—Full-Scale Room Test for at the end of high heat flux (or intensity) to the other end Surface Products). In this test, a gas burner is placed in the until they reach a location where the heat flux is not sufficient corner of the room, which has a single door for ventilation. for further propagation. This is reported as the critical radiant Three of the walls are lined with the test material, and the flux. Thus, low critical radiant flux reflects materials with ceiling may also be lined with the test material. Other high flammability. Typical requirements are for a minimum room/corner tests use a wood crib or similar item as the critical radiant flux level of 2.2 or 4.5 kW/m2 depending on ignition source. Such a room/corner test is used to regulate location and occupancy. Data in the literature indicate that foam plastic insulation, a material that is not properly evalu- oak flooring has a critical radiant flux of 3.5 kW/m2 ated in the ASTM E84 test. (Benjamin and Adams 1976). Observations are made of the growth of the fire and the dura- There is also a smoldering combustion test for cellulosic tion of the test until flashover occurs. Instruments record the insulation. Cellulosic insulation is regulated by a product heat generation, temperature development within the room, safety standard of the U.S. Consumer Product Safety Com- and the heat flux to the floor. Results of full-scale room/ mission (Interim Safety Standard for Cellulosic Insulation: corner tests are used to validate fire growth models and Cellulosic Insulation Labeling and Requirements, 44FR bench-scale test results. Fire endurance tests evaluate the 39938, 16CFR Part 1209, 1979; also Gen. Serv. Admin. relative performance of the assemblies during a post-flashover Spec. HH–I–515d). Proper chemical treatments of cellulosic fire. insulation are required to reduce its tendency for smoldering combustion and to reduce flame spread. Proper installation Containment to Compartment around recessed light fixtures and other electrical devices is necessary. of Origin The growth, intensity, and duration of the fire is the “load” Other tests for flammability include those that measure heat that determines whether a fire is confined to the room of release. Other flammability tests and fire growth modeling origin. Whether a given fire will be contained to the com- are discussed in the Fire Performance Characteristics of partment depends on the fire resistance of the walls, doors, Wood section. ceilings, and floors of the compartment. Requirements for fire resistance or fire endurance ratings of structural members and Rated roof covering materials are designated either Class A, assemblies are another major component of the building code B, or C according to their performance in the tests described provisions. Fire resistance is the ability of materials or their in ASTM E108, Fire Tests of Roof Coverings. This test assemblies to prevent or retard the passage of excessive heat, standard includes intermittent flame exposure, spread of hot gases, or flames while continuing to support their struc- flame, burning brand, flying brand, and rain tests. There is a tural loads. Fire-resistance ratings are usually obtained by different version of the pass/fail test for each of the three conducting standard fire tests. In the standard fire-resistance classes. Class A test is the most severe and Class C the test (ASTM E119), there are three failure criteria: element least. In the case of the burning brand tests, the brand for the collapse, passage of flames, or excessive temperature rise on Class B test is larger than that for the Class C test. Leach- the non-fire-exposed surface (average increase of several loca- resistant fire-retardant-treated shingles are available that carry tions exceeding 139°C or 181°C at a single location). a Class B or C fire rating. The self-insulating qualities of wood, particularly in the large Information on ratings for different products can be obtained wood sections of heavy timber construction, are an important from industry literature, evaluation reports issued by the factor in providing a degree of fire resistance. In Type IV or model code organizations, and listings published by testing heavy timber construction, the need for fire-resistance re- laboratories or quality assurance agencies. Products listed by quirements is achieved in the codes by specifying minimum Underwriters Laboratories, Inc., and other such organizations sizes for the various members or portions of a building and are stamped with the rating information. other prescriptive requirements. In this type of construction, the wood members are not required to have specific

17–4 fire-resistance ratings. The acceptance of heavy timber The relatively good structural behavior of a traditional wood construction is based on historical experience with its per- member in a fire test results from the fact that its strength is formance in actual fires. Proper heavy timber construction generally uniform through the mass of the piece. Thus, the includes using approved fastenings, avoiding concealed unburned fraction of the member retains high strength, and spaces under floors or roofs, and providing required fire its load-carrying capacity is diminished only in proportion to resistance in the interior and exterior walls. its loss of cross section. Innovative designs for structural wood members may reduce the mass of the member and In recent years, the availability and code acceptance of a locate the principal load-carrying components at the outer procedure to calculate the fire-resistance ratings for large edges where they are most vulnerable to fire, as in structural timber beams and columns have allowed their use in fire- sandwich panels. With high strength facings attached to a rated buildings not classified as heavy timber construction low-strength core, unprotected load-bearing sandwich panels (Type IV). In the other types of construction, the structural have failed to support their load in less than 6 min when members and assemblies are required to have specified fire- tested in the standard test. If a sandwich panel is to be used resistance ratings. Details on the procedure for large timbers as a load-bearing assembly, it should be protected with can be found in American Institute of Timber Construction gypsum wallboard or some other thermal barrier. In any (AITC) Technical Note 7 and the AF&PA DCA #2 “Design protected assembly, the performance of the protective mem- of Fire-Resistive Exposed Wood Members” (AWC 1985). brane is the critical factor in the performance of the assembly. The fire resistance of glued-laminated structural members, Unprotected light-frame wood buildings do not have the such as arches, beams, and columns, is approximately natural fire resistance achieved with heavier wood members. equivalent to the fire resistance of solid members of similar In these, as in all buildings, attention to good construction size. Available information indicates that laminated members details is important to minimize fire hazards. Quality of glued with phenol, resorcinol, or melamine adhesives are at workmanship is important in achieving adequate fire resis- least equal in their fire resistance to a one-piece member of tance. Inadequate nailing and less than required thickness of the same size. Laminated members glued with casein have the interior finish can reduce the fire resistance of an assem- only slightly less fire resistance. bly. The method of fastening the interior finish to the fram- Light-frame wood construction can provide a high degree of ing members and the treatment of the joints are significant fire containment through use of gypsum board as the interior factors in the fire resistance of an assembly. The type and finish. This effective protective membrane provides the initial quantity of any insulation installed within the assembly may fire resistance rating. Many recognized assemblies involving also affect the fire resistance of an assembly. Electrical recep- wood-frame walls, floors, and roofs provide a 1- or 2-hour fire tacle outlets, pipe chases, and other through openings that resistance rating. Fire-rated gypsum board (Type X or C) is are not adequately firestopped can affect the fire resistance. In used in rated assemblies. Type X and the higher grade Type addition to the design of walls, ceilings, floors, and roofs for C gypsum boards have textile glass filaments and other fire resistance, stairways, doors, and firestops are of particular ingredients that help to keep the gypsum core intact during a importance. fire. Fire-resistance ratings of various assemblies are listed in Fires in buildings can spread by the movement of hot fire the model codes and other publications such as the Fire gases through open channels in concealed spaces. Codes Resistance Design Manual (Gypsum Association). Tradi- specify where firestops and draftstops are required in con- tional constructions of regular gypsum wallboard (that is, not cealed spaces, and they must be designed to interfere with the fire rated) or lath and plaster over wood joists and studs have passage of flames up or across a building. In addition to fire-resistance ratings of 15 to 30 min. going along halls, stairways, and other large spaces, heated While fire-resistance ratings are for the entire wall, floor, or gases also follow the concealed spaces between floor joists roof assembly, the fire resistance of a wall or floor can be and between studs in partitions and walls of frame construc- viewed as the sum of the resistance of the interior finish and tion. Obstruction of these hidden channels provides an effec- the resistance of the framing members. In a code-accepted tive means of restricting fire from spreading to other parts of procedure, the fire rating of a light-frame assembly is calcu- the structure. Firestops are materials used to block off rela- lated by adding the tabulated times for the fire-exposed tively small openings passing through building components membrane to the tabulated times for the framing. For exam- such as floors and walls. Draftstops are barriers in larger ple, the fire-resistance rating of a wood stud wall with concealed spaces such as those found within wood joist floor 16-mm-thick Type X gypsum board and rock wool insula- assemblies with suspended dropped ceilings or within an tion is computed by adding the 20 min listed for the stud attic space with pitched chord trusses. wall, the 40 min listed for the gypsum board, and the Doors can be critical in preventing the spread of fires. Doors 15 min listed for the rock wool insulation to obtain a rating left open or doors with little fire resistance can easily defeat for the assembly of 75 min. Additional information on this the purpose of a fire-rated wall or partition. Listings of fire- component additive method (CAM) can be found in the rated doors, frames, and accessories are provided by various AF&PA DCA No. 4 “Component Additive Method (CAM) fire testing agencies. When a fire-rated door is selected, for Calculating and Demonstrating Assembly Fire Endur- details about which type of door, mounting, hardware, and ance” (AWC 1991). More sophisticated mechanistic models closing mechanism need to be considered. are being developed.

17–5 Fire Safety Engineering heated objects. This flow of energy or heat flux can have both convective and radiative components. The field of fire safety engineering is undergoing rapid changes because of the development of more engineering and Piloted ignition above a single flat surface has recently been scientific approaches to fire safety. This development is studied in some depth because of the advent of fire growth evidenced by the publication of The Society of Fire Protec- research. The surface temperature of wood materials has been o o tion Engineers Handbook of Fire Protection Engineering measured somewhere between 300 C to 400 C prior to pi- and formation of fire safety engineering subcommittees in loted ignition. Surface temperature at ignition is an illusive ISO and ASTM. Steady advances are being made in the quantity that is experimentally difficult to obtain. Equipment fields of fire dynamics, fire hazard calculations, fire design such as the Ohio State University (OSU) apparatus (ASTM calculations, and fire risk analysis. Such efforts support the E906), the cone calorimeter (ASTM 1354), and the lateral worldwide trend to develop alternative building codes based ignition and flame spread test (LIFT) apparatus (ASTM on performance criteria rather than prescriptive requirements. 1321) are used to obtain data on time to piloted ignition as a Additional information on fire protection can be found in the function of heater irradiance. Table 17–2 indicates the de- various publications of the National Fire Protection crease in time to ignition with the increase in imposed heat Association (NFPA). flux for different species of wood measured with the OSU apparatus. Similar, perhaps identical, materials have been tested recently in cone calorimeter and LIFT apparatuses Fire Performance with somewhat similar results. From such tests, values of ignition temperature, critical ignition flux (heat flux below Characteristics of Wood which ignition would not occur), and thermophysical proper- Wood will burn when exposed to heat and air. Thermal ties have been derived using a transient heat conduction degradation of wood occurs in stages. The degradation proc- theory. These properties are also material dependent; they ess and the exact products of thermal degradation depend depend heavily on density of the material and moisture upon the rate of heating as well as the temperatures. The content. A range of wood products tested have ignition sequence of events for wood combustion is as follows: surface temperatures of 300oC to 400oC and a critical ignition 2 • flux of between 10 and 13 kW/m in the cone calorimeter. The wood, responding to heating, decomposes or pyro- The ignition surface temperature is lower for low density lyzes into volatiles and char. Char is the dominant prod- woods. Estimates of piloted ignition in various scenarios can uct at internal temperatures less than 300oC, whereas o be obtained using the derived thermal properties and an volatiles become much more pronounced above 300 C. applicable heat conduction model. • The volatiles, some of which are flammable, can be ig- nited if the volatile–air mixture is of the right composition Some, typically old, apparatuses for testing piloted ignition in a temperature range of about 400oC to 500oC within the measured the temperature of the air flow rather than the im- mixture. This gas-phase combustion appears as flames. posed heat flux with the time to ignition measurement. These results were often reported as the ignition temperature • With air ventilation, the char oxidation becomes signifi- and as varying with time to ignition, which is misleading. cant around 200oC with two peaks in intensity reported at When the imposed heat flux is due to a radiant source, such 360oC and 520oC. This char oxidation is seen as glowing reported air flow ignition temperature can be as much as ° or smoldering combustion until only ash residue remains. 100 C lower than the ignition surface temperature. For a This solid-phase combustion will not proceed if flaming proper heat conduction analysis in deriving thermal proper- combustion prevents a supply of fresh air to the char ties, measurements of the radiant source flux and air flow rate surfaces. are also required. Since imposed heat flux to the surface and the surface ignition temperature are the factors that directly Several characteristics are used to quantify this burning determine ignition, some data of piloted ignition are behavior of wood, including ignition from heat sources, inadequate or misleading. growing rate of heat release leading to room flashover, flame spread in heated environments, smoke and toxic gases, Unpiloted ignition depends on special circumstances that flashover, and charring rates in a contained room. result in different ranges of ignition temperatures. At this time, it is not possible to give specific ignition data that Ignition apply to a broad range of cases. For radiant heating of cellu- losic solids, unpiloted transient ignition has been reported at Ignition of wood takes place when wood is subject to suffi- 600°C. With convective heating of wood, unpiloted ignition cient heat and in atmospheres that have sufficient oxygen. has been reported as low as 270°C and as high as 470°C. Ignition can be of two types: piloted or unpiloted. Piloted ignition occurs in the presence of an ignition source (such as Unpiloted spontaneous ignition can occur when a heat source a spark or a flame). Unpiloted ignition is ignition that occurs within the wood product is located such that the heat is not where no pilot source is available. The wood surface is readily dissipated. This kind of ignition involves smoldering ignited by the flow of energy or heat flux from a fire or other and generally occurs over a longer period of time. Smolder- ing is thermal degradation that proceeds without flames or

17–6 Table 17–2. Flammability data for selected wood species Effective heat of Average heat release Ignition timeb (s) Higher combustiond (MJ/kg) rateb (kW/m2) 18- 55- heating 18- 55- 18- 55- Densitya kW/m2 kW/m2 valuec kW/m2 kW/m2 kW/m2 kW/m2 Species (kg/m3) heat flux heat flux (MJ/kg) heat flux heat flux heat flux heat flux

Softwoods Pine, Southern 508 740 5 20.5 9.1 13.9 40.4 119.6 Redwood 312 741 3 21.1 10.7 14.2 39.0 85.9

Hardwoods Basswood 312 183 5 20.0 10.9 12.2 52.8 113.0 Oak, red 660 930 13 19.8 9.0 11.7 48.7 113.3 aBased on weight and volume of ovendried wood. bIgnition times, effective heat of combustion, and average rate of heat release (HRR) obtained using an ASTM E906 heat release apparatus modified to measured heat release using oxygen consumption method. Test durations were 50 to 98 min for 18-kW/m2 heat flux and 30 to 53 min for 55-kW/m2 heat flux. Test was terminated prior to the usual increase in HRR to a second peak as the specimen is consumed. cFrom oxygen bomb calorimeter test. dApparent effective heat of combustion based on average HRR and mass loss rate, which includes the moisture driven from the wood. See footnote b. visible glowing. Examples of such fires are (a) panels or Building codes do not generally regulate building materials paper removed from the press or dryer and stacked in large on the basis of ignition or ignitability. As a result, general piles without adequate cooling and (b) very large piles of fire safety design criteria have not been developed. Rather, chips or sawdust with internal exothermic reactions such as this subject is considered in conjunction with limits on biological activities. Potential mechanisms of internal heat combustibility and flame spread. generation include respiration, metabolism of microorgan- isms, heat of pyrolysis, abiotic oxidation, and adsorptive Heat Release heat. These mechanisms, often in combination, may proceed to smoldering or flaming ignition through a thermal runaway Heat release rates are important because they indicate the effect within the pile if sufficient heat is generated and is not potential fire hazard of a material and also the combustibility dissipated. The minimum environmental temperature to of a material. Materials that release their potential chemical achieve ignition is called the self-accelerating decomposition energy (and also the smoke and toxic gases) relatively temperature and includes the effects of specimen mass and air quickly are more hazardous than those that release it more ventilation. slowly. There are materials that will not pass the current definition of noncombustible in the model codes but will Unpiloted ignitions that involve wood exposed to low level release only limited amounts of heat during the initial and external heat sources over very long periods is an area of critical periods of fire exposure. There is also some criticism dispute. This kind of ignition, which involves considerable of using limited flammability to partially define noncombus- charring, does appear to occur, based on fire investigations. tibility. One early attempt was to define combustibility in However, these circumstances do not lend themselves easily terms of heat release in a potential heat method (NFPA 259), to experimentation and observation. There is some evidence with the low levels used to define low combustibility or that the char produced under low heating temperatures can noncombustibility. This test method is being used to regu- have a different chemical composition, which results in a late materials under some codes. The ground-up wood sam- somewhat lower ignition temperature than normally re- ple in this method is completely consumed during the expo- corded. Thus, a major issue is the question of safe working sure to 750°C for 2 h, which makes the potential heat for temperature for wood exposed for long periods. Temperatures ° ° wood identical to the gross heat of combustion from the between 80 C to 100 C have been recommended as safe oxygen bomb calorimeter (the higher heating value in Table surface temperatures for wood. Since thermal degradation is a 17–2). The typical gross heat of combustion averaged around prerequisite for ignition of the char layer, conservative criteria 20 MJ/kg for ovendried wood, depending on the lignin and for determining safe working temperatures can be the tem- extractive content of the wood. perature and duration needed for thermal degradation. Schaffer (1980) used a residual weight criterion of 40% of the initial A better or a supplementary measure of degrees of combusti- weight to suggest that wood can safely be heated to 150°C bility is a determination of the rate of heat release (RHR) or for a year or more before satisfying this conservative predictor heat release rate (HRR). This measurement efficiently of heating time to reach an incipient smoldering state. assesses the relative heat contribution of materials—thick,

17–7 thin, untreated, or treated—under fire exposure. The cone averaged HRR at 1-, 3-, and 5-min periods for various wood calorimeter (ASTM E1354) is the most commonly used species. bench-scale HRR apparatus and is based on the oxygen consumption method. An average value of 13.1 kJ/g of Heat release rate depends upon the intensity of the imposed oxygen consumed was the constant found for organic solids heat flux. Table 17–2 provides the average effective heat of combustion and average HRR for four wood species and two and is accurate with very few exceptions to within 5%. Thus, 2 it is sufficient to measure the mass flow rate of oxygen con- levels of heat flux (18 and 55 kW/m ). These results were sumed in a combustion system to determine the net HRR. obtained in an OSU apparatus modified by the Forest Prod- The procedure known as ASTM E906 (the OSU apparatus) ucts Laboratory (FPL). Similar values were also obtained in is a well-known and widely used calorimeter based on meas- the cone calorimeter (Table 17–3). Generally, the averaged urements of heat content of incoming and exiting air flow effective heat of combustion is about 65% of the oxygen through the apparatus. Because of the errors caused by the bomb heat of combustion (higher heating value) with a small heat losses and the fact that the mass flow rate is controlled linear increase with irradiance. The HRR itself has a large in the OSU apparatus, several researchers have modified it to linear increase with the heat flux. Data indicate that HRRs the oxygen consumption method. These bench-scale appara- decrease with increasing moisture content of the sample and tuses use a radiant source to provide the external heat expo- are markedly reduced by fire-retardant treatment (Fig. 17–1). sure to the test specimen. The imposed heat flux is kept constant at a specified heat flux level. The intermediate-scale Flame Spread apparatus (ASTM E1623) for testing 1- by 1-m assemblies or composites and the room full-scale test (ISO 9705) also The spread of flames over solids is a very important phe- use the oxygen consumption technique to measure the HRR nomenon in the growth of compartment fires. Indeed, in fires of fires at larger scales. where large fuel surfaces are involved, the increase in HRR with time is primarily due to the increase in burning area. The cone calorimeter is ideal for product development with Many data have been acquired with the flame spread tests its small specimen size of 100 by 100 mm. The specimen is used in building codes. Table 17–1 lists the FSI and smoke continuously weighed by use of a load cell. In conjunction index of ASTM E84 for solid wood. Some consistencies in with HRR measurements, the effective heat of combustion as the FSI behavior of the hardwood species can be related to a function of time is calculated by the ASTM E1354 their density. Considerable variations are found for wood- method. Basically, the effective heat of combustion is the based composites; for example, the FSI of four structural HRR divided by the mass loss rate as determined from the flakeboards ranged from 71 to 189. cone calorimeter test as a function of time. A typical HRR profile as shown in Figure 17–1 for plywood begins with a As a prescriptive regulation, the ASTM E84 tunnel test is a sharp peak upon ignition, and as the surface chars, the HRR success in the reduction of fire hazards but is impractical in drops to some minimum value. After the thermal wave providing scientific data for fire modeling or in useful bench- travels completely through the wood thickness, the back side scale tests for product development. Other full-scale tests of a wood sample reaches pyrolysis temperature, thus giving (such as the ISO 9705 room/corner test) also use both an rise to a second, broader, and even higher HRR peak. For ignition burner and the ensuing flame spread to assist flow fire-retardant-treated wood products, the first HRR peak may but can produce quite different results because of the size of be reduced or eliminated. Table 17–3 provides the peak and the ignition burner or the test geometry. This is the case with foam plastic panels that melt and drip during a fire test. In the tunnel test, with the test material on top, a material that melts can have low flammability since the specimen 200 does not stay in place. With an adequate burner in the FRT room/corner test, the same material will exhibit very high Untreated flammability. 150 )

2 A flame spreads over a solid material when part of the fuel, ahead of the pyrolysis front, is heated to the critical condition of ignition. The rate of flame spread is controlled by how 100 rapidly the fuel reaches the ignition temperature in response to heating by the flame front and external sources. The mate- HRR (kW/m rial’s thermal conductivity, heat capacitance, thickness, and 50 blackbody surface reflectivity influence the material’s thermal response, and an increase in the values of these properties corresponds to a decrease in flame spread rate. On the other hand, an increase in values of the flame features, such as the 0 200 400 600 800 imposed surface fluxes and spatial lengths, corresponds to a Time (s) increase in the flame spread rate. Figure 17–1. Heat release curves for untreated and FRT plywood exposed to 50-kW/m2 radiance.

17–8 Table 17–3. Heat release data for selected wood speciesa Average effective Densityb Heat release rate (kW/m2) heat of combustionc Ignition Species (kg/m3) Peak 60-s avg 180-s avg 300-s avg (MJ/kg) time (s) Softwoods Pine, red 525 209 163 143 132 12.9 24 Pine, white 359 209 150 117 103 13.6 17 Redcedar, eastern — 175 92 95 85 11.7 25 Redwood 408 227 118 105 95 13.2 17

Hardwoods Birch 618 218 117 150 141 12.2 29 Maple, hard 626 218 128 146 137 11.7 31 Oak, red 593 214 115 140 129 11.4 28 aData for 50-kW/m2 heat flux in cone calorimeter. Tested in specimen holder without retaining frame. Specimens conditioned to 23°C, 50% relative humidity. bOvendry mass and volume. cTests terminated when average mass loss rate dropped below 1.5 g/s m2 during 1-min period.

Flame spread occurs in different configurations, which are Smoke and Toxic Gases organized by orientation of the fuel and direction of the main flow of gases relative to that of flame spread. Downward and One of the most important problems associated with fires is lateral creeping flame spread involves a fuel orientation with the smoke they produce. The term smoke is frequently used buoyantly heated air flowing opposite of the flame spread in an all-inclusive sense to mean the mixture of pyrolysis direction. Related bench-scale test methods are ASTM E162 products and air that is present near the fire site. In this for downward flame spread, ASTM E648 for horizontal flame context, smoke contains gases, solid particles, and droplets spread to the critical flux level, and ASTM E1321 (LIFT of liquid. Smoke presents potential hazards because it inter- apparatus) for lateral flame spread on vertical specimen to the acts with light to obscure vision and because it contains critical flux level. The heat transfer from the flame to the noxious and toxic substances. virgin fuel is primarily conductive within a spatial extent of a few millimeters and is affected by ambient conditions such as Generally, two approaches are used to deal with the smoke oxygen, pressure, buoyancy, and external irradiance. For problem: limit smoke production and control the smoke that most wood materials, this heat transfer from the flame is less has been produced. The control of smoke flow is most often than or equal to surface radiant heat loss in normal ambient a factor in the design and construction of large or tall build- conditions, so that excess heat is not available to further raise ings. In these buildings, combustion products may have the virgin fuel temperature; flame spread is prevented as a serious effects in areas remote from the actual fire site. result. Therefore, to achieve creeping flame spread, an exter- nal heat source is required in the vicinity of the pyrolysis Currently, several bench-scale test methods provide compara- front. tive smoke yield information on materials and assemblies. Each method has entirely different exposure conditions; none Upward or ceiling flame spread involves a fuel orientation is generally correlated to full-scale fire conditions or experi- with the main air flowing in the same direction as the flame ence. Until the middle 1970s, smoke yield restrictions in spread (assisting flow). At present, there are no small-scale building codes were almost always based on data from tests for upward flame spread potential. Thus, testing of ASTM E84. The smoke measurement is based on a percent- flame spread in assisting flow exists mostly in both the age attenuation of white light passing through the tunnel tunnel tests and the room/corner burn tests. The heat transfer exhaust stream and detected by a photocell. This is con- from the flame is both conductive and radiative, has a large verted to the smoke development index (SDI), with red oak spatial feature, and is relatively unaffected by ambient condi- flooring set at 100. The flame spread requirements for interior tions. Rapid acceleration in flame spread can develop because finish generally are linked to an added requirement that the of a large, increasing magnitude of flame heat transfer as a SDI be less than 450. result of increasing total HRR in assisting flows. These complexities and the importance of the flame spread proc- In the 1970s, the apparatus known as the NBS smoke cham- esses explain the many and often incompatible flame spread ber was developed and approved as an ASTM standard for tests and models in existence worldwide. research and development (ASTM E662). This test is a static smoke test because the specimen is tested in a closed

17–9 chamber of fixed volume and the light attenuation is recorded of test furnace and the adjustment methods used in a stan- over a known optical path length. The corresponding light dardized toxicity test. transmission is reported as specific optical density as a func- tion of time. Samples are normally tested in both flaming Charring and Fire Resistance (pilot flame) and nonflaming conditions using a radiant flux of 25 kW/m2. As noted earlier in this chapter, wood exposed to high tem- peratures will decompose to provide an insulating layer of The dynamic measurement of smoke in the heat release char that retards further degradation of the wood. The load- calorimeter (ASTM E906 and E1354) has recently gained carrying capacity of a structural wood member depends upon increasing recognition and use. The E906 and E1354 tests its cross-sectional dimensions. Thus, the amount of charring are dynamic in that the smoke continuously flows out the of the cross section is the major factor in the fire endurance of exhaust pipe where the optical density is measured continu- structural wood members. ously. The appropriate smoke parameter is the smoke release When wood is first exposed to fire, the wood chars and rate (SRR), which is the optical density multiplied by the eventually flames. Ignition occurs in about 2 min under the volume flow rate of air into the exhaust pipe and divided by standard ASTM E119 fire-test exposures. Charring into the the product of exposed surface area of the specimen and the depth of the wood then proceeds at a rate of approximately light path length. Often the smoke extinction area, which is 0.8 mm/min for the next 8 min (or 1.25 min/mm). There- the product of SRR and the specimen area, is preferred be- after, the char layer has an insulating effect, and the rate cause it can be correlated linearly with HRR in many cases. decreases to 0.6 mm/min (1.6 min/mm). Considering the This also permits comparison with the smoke measured in initial ignition delay, the fast initial charring, and then the the room/corner fire test because HRR is a readily available slowing down to a constant rate, the average constant char- test result. Although SRR can be integrated with time to get ring rate is about 0.6 mm/min (or 1.5 in/h) (Douglas-fir, the same units as the specific optical density, they are not 7% moisture content). In the standard fire-resistance test, equivalent because static tests involve the direct accumula- this linear charring rate is generally assumed for solid wood tion of smoke in a volume, whereas SRR involves accumula- directly exposed to fire. tion of freshly entrained air volume flow for each unit of smoke. Methods investigated to correlate smoke between There are differences among species associated with their different tests included alternative parameters such as density, anatomy, chemical composition, and permeability. particulate mass emitted per area of exposed sample. Moisture content is a major factor affecting charring rate. Density relates to the mass needed to be degraded and the Toxicity of combustion products is an area of concern. About thermal properties, which are affected by anatomical features. 75% to 80% of fire victims are not touched by flame but die Charring in the longitudinal grain direction is reportedly as a result of exposure to smoke, exposure to toxic gases, or double that in the transverse direction, and chemical compo- oxygen depletion. These life-threatening conditions can sition affects the relative thickness of the char layer. Perme- result from burning contents, such as furnishings, as well as ability affects the movement of moisture being driven from from the structural materials involved. The toxicity resulting the wood or that being driven into the wood beneath the char from the thermal decomposition of wood and cellulosic layer. Normally, a simple linear model for charring where t is substances is complex because of the wide variety of types of time (min), C is char rate (min/mm), and xc is char depth wood smoke. The composition and the concentration of the (mm) is assumed: individual constituents depend on such factors as the fire exposure, the oxygen and moisture present, the species of tCx = c (17–1) wood, any treatments or finishes that may have been applied, and other considerations. Toxicity data may be more widely The temperature at the base of the char layer is generally available in the future with the recent adoption of a standard taken to be 300°C or 550°F (288°C). With this temperature test method (ASTM E1678). criterion, empirical equations for charring rate have been developed. Equations relating charring rate under ASTM Carbon monoxide is a particularly insidious toxic gas. Small E119 fire exposure to density and moisture content are avail- amounts of carbon monoxide are particularly toxic because able for Douglas-Fir, Southern Pine, and White Oak. These the hemoglobin in the blood is much more likely to com- equations for rates transverse to the grain are bine with carbon monoxide than with oxygen, even with µ ρ plenty of breathable oxygen. This poisoning is called car- C = (0.002269 + 0.00457 ) + 0.331 for Douglas Fir boxyhemoglobin. Recent research has shown that the kind of (17–2a) fires that kill people by toxicity are principally those that C = (0.000461 + 0.00095µ)ρ + 1.016 for Southern Pine reach flashover in a compartment or room some distance from (17–2b) the people. The vast majority of fires that attain flashover µ ρ generate dangerous levels of carbon monoxide, independent C = (0.001583 + 0.00318 ) + 0.594 for White Oak of what is burning. The supertoxicants, such as hydrogen (17–2c) cyanide and neurotoxin, have been proven to be extremely where µ is moisture content (fraction of ovendry mass) and rare, even in the laboratory. These factors impact the choice ρ is density, dry mass volume at moisture content µ (kg/m3).

17–10 Table 17–4. Charring rate data for selected wood species Wood exposed to ASTM E119 exposurea Wood exposed to a constant heat fluxb Linear charring ratee Thermal penetra- Average mass loss Non- (min/mm) tion depth d g ( mm) rate (g/m2 s) Linear linear Thermal Char charring charring penetra- 55- 18- 55- 55- Den- contrac- ratee ratef tion 18 - kW/m2 kW/m2 kW/m2 18- kW/m2 sityc tion (min/ (min/ depth g kW/m2 heat heat heat kW/m2 heat Species (kg/m3) factord mm) mm1.23) (mm) heat flux flux flux flux heat flux flux Softwoods Southern Pine 509 0.60 1.24 0.56 33 2.27 1.17 38 26.5 3.8 8.6 Western 310 0.83 1.22 0.56 33 — — — — — — redcedar Redwood 343 0.86 1.28 0.58 35 1.68 0.98 36.5 24.9 2.9 6.0 Engelmann 425 0.82 1.56 0.70 34 — — — — — — spruce

Hardwoods Basswood 399 0.52 1.06 0.48 32 1.32 0.76 38.2 22.1 4.5 9.3 Maple, hard 691 0.59 1.46 0.66 31 — — — — — — Oak, red 664 0.70 1.59 0.72 32 2.56 1.38 27.7 27.0 4.1 9.6 Yellow- 504 0.67 1.36 0.61 32 — — — — — — poplar aMoisture contents of 8% to 9%. bCharring rate and average mass loss rate obtained using ASTM E906 heat release apparatus. Test durations were 50 to 98 min for 18-kW/m2 heat flux and 30 to 53 min for 55-kW/m2 heat flux. Charring rate based on temperature criterion of 300°C and linear model. Mass loss rate based on initial and final weight of sample, which includes moisture driven from the wood. Initial average moisture content of 8% to 9%. cBased on weight and volume of ovendried wood. dThickness of char layer at end of fire exposure divided by original thickness of charred wood layer (char depth). eBased on temperature criterion of 288°C and linear model. fBased on temperature criterion of 288°C and nonlinear model of Equation (17–3). gAs defined in Equation (17–6). Not sensitive to moisture content.

A nonlinear char rate model has been found useful. This Charring rate is also affected by the severity of the fire expo- alternative model is sure. Data on charring rates for fire exposures other than ASTM E119 have been limited. Data for exposure to con- 1.23 ° ° ° tmx = c (17–3) stant temperatures of 538 C, 815 C, and 927 C are available 1.23 in Schaffer (1967). Data for a constant heat flux are given in where m is char rate coefficient (min/mm ). Table 17–4. Based on data from eight species (Table 17–4), the following The temperature at the innermost zone of the char layer is equation was developed for the char rate coefficient: assumed to be 300°C. Because of the low thermal conductiv- − ρ µ ity of wood, the temperature 6 mm inward from the base of m = 0.147 + 0.000564 + 1.21 + 0.532 fc (17–4) the char layer is about 180°C. This steep temperature gradi- ρ ent means the remaining uncharred cross-sectional area of a where is density, ovendry mass and volume, and fc is char contraction factor (dimensionless). large wood member remains at a low temperature and can continue to carry a load. Moisture is driven into the wood as The char contraction factor is the thickness of the residual charring progresses. A moisture content peak is created inward from the char base. The peak moisture content occurs char layer divided by the original thickness of the wood layer ° that was charred (char depth). Average values for the eight where the temperature of the wood is about 100 C, which is species tested in the development of the equation are listed in at about 13 mm from the char base. Table 17–4. Once a quasi-steady-state charring rate has been obtained, the These equations and data are valid when the member is thick temperature profile beneath the char layer can be expressed as enough to be a semi-infinite slab. For smaller dimensions, an exponential term or a power term. An equation based on a the charring rate increases once the temperature has risen power term is above the initial temperature at the center of the member or at 2 T = T +()300 − T() 1− x/d (17–5) the unexposed surface of the panel. As a beam or column ii chars, the corners become rounded.

17–11 ° ° where T is temperature ( C), Ti initial temperature ( C), While fire-retardant-treated wood is not considered a non- x distance from the char front (mm), and d thermal penetra- combustible material, many codes have accepted the use of tion depth (mm). fire-retardant-treated wood and plywood in fire-resistive and noncombustible construction for the framing of nonload- In Table 17–4, values for the thermal penetration depth bearing walls, roof assemblies, and decking. Fire-retardant- parameter are listed for both the standard fire exposure and treated wood is also used for such special purposes as wood the constant heat flux exposure. As with the charring rate, scaffolding and for the frame, rails, and stiles of wood fire these temperature profiles assume a semi-infinite slab. The doors. equation does not provide for the plateau in temperatures that often occurs at 100°C in moist wood. In addition to these In addition to specifications for flame spread performance, empirical data, there are mechanistic models for estimating fire-retardant-treated wood for use in certain applications is the charring rate and temperature profiles. The temperature specified to meet other performance requirements. Wood profile within the remaining wood cross-section can be used treated with inorganic flame-retardant salts is usually more with other data to estimate the remaining load-carrying hygroscopic than is untreated wood, particularly at high capacity of the uncharred wood during a fire and the residual relative humidities. Increases in equilibrium moisture con- capacity after a fire. tent of this treated wood will depend upon the type of chemi- cal, level of chemical retention, and size and species of wood involved. Applications that involve high humidity will Flame-Retardant Treatments likely require wood with low hygroscopicity. The American To meet building code and standards specifications, lumber Wood Preservers’ Association (AWPA) Standards C20 and and plywood are treated with flame retardants to improve C27 requirements for low hygroscopicity (Interior Type A their fire performance. The two general application methods treatment) stipulate that the material shall have an equilib- are pressure treating and surface coating. rium moisture content of not more than 28% when tested in accordance with ASTM D3201 procedures at 92% relative humidity. Fire-Retardant-Treated Wood To meet the specifications in the building codes and various Exterior flame-retardant treatments should be specified when- standards, fire-retardant-treated lumber and plywood is wood ever the wood is exposed to exterior weathering conditions. that has been pressure treated with chemicals to reduce its The AWPA Standards C20 and C27 also mandate that an flame spread characteristics. Flame-retardant treatment of exterior type treatment is one that has shown no increase in wood generally improves the fire performance by reducing the fire hazard classification after being subjected to the rain test amount of flammable volatiles released during fire exposure specified in ASTM D2898 as Method A. or by reducing the effective heat of combustion, or both. For structural applications, information on the fire-retardant- Both results have the effect of reducing the HRR, particularly treated wood product needs to be obtained from the treater or during the initial stages of fire, and thus consequently chemical supplier. This includes the design modification reducing the rate of flame spread over the surface. The wood factors for initial strength properties of the fire-retardant- may then self-extinguish when the primary heat source is treated wood, including values for the fasteners. Flame- removed. retardant treatment generally results in reductions in the The performance requirement for fire-retardant-treated wood is mechanical properties of wood. Fire-retardant-treated wood is that its FSI is 25 or less when tested according to the often more brash than untreated wood. ASTM E84 flame spread test and that it shows no evidence In field applications with elevated temperatures, such as roof of significant progressive combustion when this 10-min test sheathings, there is the potential for further losses in strength is continued for an additional 20 min. In addition, it is with time. For such applications in elevated temperatures required that the flame front in the test shall not progress and high humidity, appropriate design modification factors more than 3.2 m beyond the centerline of the burner at any need to be obtained from the treater or chemical supplier. given time during the test. Underwriters Laboratories, Inc., The AWPA Standards C20 and C27 mandate that fire- assigns the designation FR–S to products that satisfy these retardant-treated wood that will be used in high-temperature requirements. In applications where the requirement is not for applications (Interior Type A High Temperature), such as fire-retardant-treated wood but only for Class I or II flame roof framing and roof sheathing, be strength tested in accor- spread, the flame-retardant treatments only need to reduce the dance with ASTM D5664 (lumber) or ASTM D5516 FSI to the required level in the ASTM E84 flame spread test (plywood) or by an equivalent methodology. Some flame- (25 for Class I, 75 for Class II). Various laboratories perform retardant treatments are not acceptable because of thermal fire-performance rating tests on these treated materials and degradation of the wood that will occur with time at high maintain lists of products that meet certain standards. temperatures. Screw-withdrawal tests to predict residual in-place strength of fire-retardant-treated plywood roof sheath- Fire-retardant-treated wood and plywood are often used for ing have been developed (Winandy and others 1998). interior finish and trim in rooms, auditoriums, and corridors where codes require materials with low surface flammability. Corrosion of fasteners can be accelerated under conditions of high humidity and in the presence of flame-retardant salts.

17–12 For flame-retardant treatments containing inorganic salts, the Flame-Retardant Coatings type of metal and chemical in contact with each other greatly affects the rate of corrosion. Thus, information on proper For some applications, the alternative method of applying fasteners also needs to be obtained from the treater or chemi- the flame-retardant chemical as a coating to the wood surface cal supplier. Other issues that may require contacting the may be acceptable. Such commercial coating products are treater or chemical supplier include machinability, gluing available to reduce the surface flammability characteristics of characteristics, and paintability. wood. The two types of coatings are intumescent and nonintumescent. The widely used intumescent coatings Flame-retardant treatment of wood does not prevent the wood “intumesce” to form an expanded low-density film upon from decomposing and charring under fire exposure (the rate exposure to fire. This multicellular carbonaceous film insu- of fire penetration through treated wood approximates the rate lates the wood surface below from the high temperatures. through untreated wood). Fire-retardant-treated wood used in Intumescent formulations include a dehydrating agent, a char doors and walls can slightly improve fire endurance of these former, and a blowing agent. Potential dehydrating agents doors and walls. Most of this improvement is associated include polyammonium phosphate. Ingredients for the char with the reduction in surface flammability rather than any former include starch, glucose, and dipentaerythritol. Poten- changes in charring rates. tial blowing agents for the intumescent coatings include urea, melamine, and chlorinate parafins. Nonintumescent coating Flame-Retardant Pressure Treatments products include formulations of the water-soluble salts such as diammonium phosphate, ammonium sulfate, and borax. In the impregnation treatments, wood is pressure impreg- nated with chemical solutions using pressure processes similar to those used for chemical preservative treatments. References However, considerably heavier absorptions of chemicals are necessary for flame-retardant protection. Standards C20 and General C27 of the AWPA recommend the treating conditions for APA—The Engineered Wood Association. [Current lumber and plywood. The penetration of the chemicals into edition]. Fire-rated systems. Tacoma, WA: APA—The the wood depends on the species, wood structure, and mois- Engineered Wood Association. ture content. Since some species are difficult to treat, the degree of impregnation needed to meet the performance re- Browne, F.L. 1958. Theories of the combustion of wood quirements for fire-retardant-treated wood may not be possi- and its control—a survey of the literature. Rep. No. 2136. ble. One option is to incise the wood prior to treatment to Madison, WI: U.S. Department of Agriculture, Forest improve the depth of penetration. Service, Forest Products Laboratory. Inorganic salts are the most commonly used flame retardants CWC. 1996. Fire safety design in buildings. Ottawa, ON, for interior wood products, and their characteristics have Canada: Canadian Wood Council. been known for more than 50 years. These salts include monoammonium and diammonium phosphate, ammonium NFPA. 1995. Guide for fire and explosion investigations. sulfate, zinc chloride, sodium tetraborate, and boric acid. NFPA 921. Quincy, MA: National Fire Protection Guanylurea phosphate is also used. These chemicals are Association. combined in formulations to develop optimum fire perform- ance yet still retain acceptable hygroscopicity, strength, NFPA. [Current edition]. Fire protection handbook. Quincy, corrosivity, machinability, surface appearance, glueability, MA: National Fire Protection Association. and paintability. Cost is also a factor in these formulations. Schaffer, E.L.; White, R.H.; Brenden, J. 1989. Part II. Many commercial formulations are available. The AWPA Fire safety. In: Light-frame wall and floor systems—analysis Standard P17 provides information on formulations of some and performance. Gen. Tech. Rep. FPL–GTR–59. Madison, current proprietary waterborne treatments. The fire-retardant WI: U.S. Department of Agriculture, Forest Service, Forest salts are water soluble and are leached out in exterior applica- Products Laboratory: 51–86. tions or with repeated washings. Water-insoluble organic flame retardants have been developed to meet the need for Society of Fire Protection Engineers. [Current edition]. leach-resistant systems. Such treatments are also an alterna- The Society of Fire Protection Engineers handbook of fire tive when a low hygroscopic treatment is needed. These protection engineering. Quincy, MA: National Fire water-insoluble systems include (a) resins polymerized after Protection Association. impregnation into wood and (b) graft polymer flame retar- dants attached directly to cellulose. An amino resin system based on urea, melamine, dicyandiamide, and related compounds is of the first type.

17–13 Fire Test Standards Underwriters Laboratories, Inc. [Current edition]. Build- ing materials directory. Northbrook, IL: Underwriters ASTM. [Current edition]. West Conshohocken, PA: Ameri- Laboratories, Inc. can Society for Testing and Materials. Underwriters Laboratories, Inc. [Current edition]. Fire ASTM E84. Surface burning characteristics of building resistance directory. Northbrook, IL: Underwriters Laborato- materials. ries, Inc. ASTM E108. Fire tests of roof coverings. U.S. Department of Housing and Urban Development. ASTM E119. Fire tests of building construction and 1980. Guideline on fire ratings of archaic materials and as- materials. semblies. Rehabilitation Guidelines. Part 8. Washington, DC: Superintendent of Documents. ASTM E136. Behavior of materials in a vertical tube ° furnace at 750 C. Ignition ASTM E162. Surface flammability of materials using a Brenden, J.J.; Schaffer, E. 1980. Smoldering wave-front radiant heat energy source. velocity in fiberboard. Res. Pap. FPL 367. Madison, WI: ASTM E648. Critical radiant flux of floor-covering U.S. Department of Agriculture, Forest Service, Forest systems using a radiant heat energy source. Products Laboratory. ASTM E662. Specific optical density of smoke generated Dietenberger, M.A. Ignitability analysis of siding materials by solid materials. using modified protocol for LIFT apparatus. In: Proceedings, 3d Fire and Materials Conference; 1994 October 27–28; ASTM E906. Heat and visible smoke release rates from Crystal City, VA. London: Interscience Communications materials and products. Limited: 259–268. ASTM E970. Critical radiant flux of exposed attic floor Kubler, H. 1990. Self-heating of lignocellulosic materials. insulation using a radiant heat energy source. In: Nelson, G.L., ed. Fire and polymers—hazards identifica- ASTM E1321. Determining material ignition and flame tion and prevention. ACS Symposium Series 425. Washing- spread properties. ton DC: American Chemical Society: 429–449. ASTM E1354. Heat and visible smoke release rates for LeVan, S.; Schaffer, E. 1982. Predicting weight loss from materials and products using an oxygen consumption smoldering combustion in cellulosic insulation. Journal of calorimeter. Thermal Insulation. 5: 229–244. ASTM E1623. Determination of fire and thermal parame- Matson, A.F.; Dufour, R.E.; Breen, J.F. 1959. Part II. ters of materials, products, and systems using an interme- Survey of available information on ignition of wood exposed diate scale calorimeter (ICAL). to moderately elevated temperatures. In: Performance of type B gas vents for gas-fired appliances. Bull. of Res. 51. ASTM E1678. Measuring smoke toxicity for use in fire Chicago, IL: Underwriters Laboratories: 269–295. hazard analysis. Shafizadeh, F.; Sekiguchi, Y. 1984. Oxidation of chars ICBO. [Current edition]. Fire tests of door assemblies. during smoldering combustion of cellulosic materials. Uniform Building Code Standard 7-2. Whittier, CA: Interna- Combustion and Flame. 55: 171–179. tional Conference of Building Officials. Schaffer, E.L. 1980. Smoldering in cellulosics under ISO. Fire tests—full-scale room test for surface products. prolonged low-level heating. Fire Technology. 16(1): 22–28. ISO 9705. International Organization for Standardization, Geneva, Switzerland. Flame Spread NFPA. [Current edition]. Potential heat of building materi- als. NFPA 259. Quincy, MA: National Fire Protection AWC. 1999. Flame spread performance of wood products Association. (1998). Design for code acceptance No. 1. http://www.forestprod.org/awcpubs.html (11 Feb. 1999). Fire-Rated Products and Assemblies Benjamin, I.A.; Adams, C.H. 1976. The flooring radiant American Insurance Association. [Current edition]. panel test and proposed criteria. Fire Journal. 70(2): 63–70. Fire resistance ratings. New York: American Insurance March. Association. Holmes, C.A.; Eickner, H.W.; Brenden, J.J.; White, Gypsum Association. [Current edition]. Fire resistance R.H. 1979. Fire performance of structural flakeboard from design manual. Washington, DC: Gypsum Association. forest residue. Res. Pap. FPL–RP–315. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Technomic Publishing Co., Inc. 1994. Handbook of fire- Laboratory. retardant coatings and fire testing services. Lancaster, PA: Technomic Publishing Co., Inc.

17–14 Underwriters Laboratories, Inc. 1971. Wood-fire hazard White, R.H. 1987. Effect of lignin content and extractives classification. Card Data Service, Serial No. UL527, on the higher heating value of wood. Wood and Fiber Chicago, IL: Underwriters Laboratories, Inc. Science. 19(4): 446–452. Yarbrough, D.W.; Wilkes, K.E. 1997. Thermal properties and use of cellulosic insulation produced from recycled paper. Smoke and Other In: The use of recycled wood and paper in building applica- Combustion Products tions. Proc. 7286. Madison, WI: Forest Products Society: Brenden, J.J. 1970. Determining the utility of a new opti- 108–114. cal test procedure for measuring smoke from various wood products. Res. Pap. FPL 137. Madison, WI: U.S. Depart- Flashover and Room/Corner Tests ment of Agriculture, Forest Service, Forest Products Bruce, H.D. 1959. Experimental dwelling—room fires. Laboratory. Rep. 1941. Madison, WI: U.S. Department of Agriculture, Brenden, J.J. 1975. How nine inorganics salts affected Forest Service, Forest Products Laboratory. smoke yield from Douglas-fir plywood. Res. Pap. FPL 249. Dietenberger, M.A.; Grexa, O.; White, R.H. [and oth- Madison, WI: U.S. Department of Agriculture, Forest ers]. 1995. Room/corner test of wall linings with 100/300 Service, Forest Products Laboratory. kW burner. In: Proceedings, 4th international fire and mate- Hall, J.R., Jr. 1996. Whatever happened to combustion rials conference; 1995 November 15–16; Crystal City, MD. toxicity. Fire Technology. 32(4): 351–371. London: InterScience Communications Limited: 53–62. Tran, H.C. 1990. Correlation of wood smoke produced Holmes, C. 1978. Room corner-wall fire tests of some struc- from NBS smoke chamber and OSU heat release apparatus. tural sandwich panels and components. Journal of Fire & In: Hasegawa, H.K. ed., Characterization and toxicity of Flammability. 9: 467–488. smoke. ASTM STP 1082. Philadelphia, PA: American Holmes, C.; Eickner, H.; Brenden, J.J. [and others]. Society for Testing and Materials: 135–146. 1980. Fire development and wall endurance in sandwich and wood-frame structures. Res. Pap. FPL 364. Madison, WI: Charring and Fire Resistance U.S. Department of Agriculture, Forest Service, Forest AITC. [Current edition]. Calculation of fire resistance of Products Laboratory. glued laminated timbers. Tech. Note 7. Englewood, CO: Schaffer, E.L.; Eickner, H.W. 1965. Effect of wall linings American Institute of Timber Construction. on fire performance within a partially ventilated corridor. ASCE. 1982. Evaluation, maintenance and upgrading of Res. Pap. FPL 49. Madison, WI: U.S. Department of wood structures—a guide and commentary. New York: Agriculture, Forest Service, Forest Products Laboratory. American Society of Civil Engineers. 428 p. Tran, H.C. 1991. Wall and corner fire tests on selected AWC. 1985. Design of fire-resistive exposed wood members wood products. Journal of Fire Sciences. 9: 106–124. (1985). Design for code acceptance No. 2. Pub. T18 Wash- March/April. ington, DC: American Wood Council. Heat Release and Heat of Combustion AWC. 1991. Component additive method (CAM) for calcu- lating and demonstrating assembly fire endurance. DCA No. Babrauskas, V.; Grayson, S.J. eds. 1992. Heat release in 4, Design for code acceptance. Pub. T20. Washington, DC: fires. New York: Elsevier Applied Science. American Forest & Paper Association. Chamberlain, D.L. 1983. Heat release rates of lumber and Janssens, M. 1994. Thermo-physical properties for wood wood products. In: Schaffer, E.L., ed., Behavior of polymeric pyrolysis models. In: Proceedings, Pacific Timber Engineer- materials in fire. ASTM STP 816. Philadelphia, PA: ing conference; 1994 July 11–15; Gold Coast, Australia. American Society for Testing and Materials: 21–41. Fortitude Valley MAC, Queensland, Australia: Timber Tran, H.C. 1990. Modifications to an Ohio State Univer- Research and Development Advisory Council: 607–618. sity apparatus and comparison with cone calorimeter results. Janssens, M.L.; White, R.H. 1994. Short communication: In: Quintiere, J.G.; Cooper, L.Y. eds. Heat and mass transfer temperature profiles in wood members exposed to fire. Fire in fires. Proceedings, AIAA/ASME thermophysics and heat and Materials. 18: 263–265. transfer conference; 1990 June 18–20; Seattle, WA. New York: The American Society of Mechanical Engineers: Schaffer, E.L. 1966. Review of information related to the 141: 131–139. charring rate of wood. Res. Note FPL–145. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Tran, H.C. 1992. (B) Experimental data on wood materials. Products Laboratory. In: Babrauskas, V.; Grayson, S.J. eds. Heat release in fires. Chapter 11 Part. B. New York: Elsevier Applied Science: Schaffer, E.L. 1967. Charring rate of selected woods- 357–372. transverse to grain. Res. Pap. FPL 69. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory.

17–15 Schaffer, E.L. 1977. State of structural timber fire endur- Treated Wood Literature ance. Wood and Fiber. 9(2): 145–170. Holmes, C.A. 1977. Effect of fire-retardant treatments on Schaffer, E.L. 1984. Structural fire design: Wood. Res. performance properties of wood. In: Goldstein, I.S., ed. Pap. FPL 450. Madison, WI: U.S. Department of Agricul- Wood technology: Chemical aspects. Proceedings, ACS ture, Forest Service, Forest Products Laboratory. symposium Series 43. Washington, DC: American Tran, H.C.; White, R.H. 1992. Burning rate of solid wood Chemical Society. measured in a heat release calorimeter. Fire and Materials. LeVan, S.L. 1984. Chemistry of fire retardancy. In: Rowell, 16: 197–206. Roger M., ed. The chemistry of solid wood. Advances in White, R.H. 1995. Analytical methods for determining fire Chemistry Series 207. Washington, DC: American Chemi- resistance of timber members. In: The SFPE handbook of fire cal Society. protection engineering. 2d ed. Quincy, MA: National Fire LeVan, S.L.; Holmes, C.A. 1986. Effectiveness of fire- Protection Association. retardant treatments for shingles after 10 years of outdoor White, R.H.; Nordheim, E.V. 1992. Charring rate of wood weathering. Res. Pap. RP–FPL–474. Madison, WI: U.S. for ASTM E119 exposure. Fire Technology. 28: 5–30. Department of Agriculture, Forest Service, Forest Products White, R.H.; Tran, H.C. 1996. Charring rate of wood Laboratory. exposed to a constant heat flux. In: Proceedings, Wood & LeVan, S.L.; Tran, H.C. 1990. The role of boron in flame- Fire Safety, third international conference. The High Tatras; retardant treatments. In: Hamel, Margaret, ed. 1st Interna- 1996 May 6–9; Zvolen, Slovakia. Zvolen, Slovakia: Fac- tional conference on wood protection with diffusible preserva- ulty of Wood Technology, Technical University: 175–183. tives: Proceedings 47355; 1990 November 28–30; Nashville, Woeste, F.E.; Schaffer, E.L. 1981. Reliability analysis of TN. Madison, WI: Forest Products Research Society: fire-exposed light-frame wood floor assemblies. Res. Pap. 39–41. FPL 386. Madison, WI: U.S. Department of Agriculture, LeVan, S.L.; Winandy, J.E. 1990. Effects of fire-retardant Forest Service, Forest Products Laboratory. treatments on wood strength: a review. Wood and Fiber Science. 22(1): 113–131. Fire-Retardant-Treated Wood LeVan, S.L.; Ross, R.J.; Winandy, J.E. 1990. Effects of Treated Wood Standards fire retardant chemicals on the bending properties of wood at ASTM. [Current edition]. West Conshohocken, PA: elevated temperatures. Res. Pap. FPL–RP–498. Madison, American Society for Testing and Materials. WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory. ASTM E69. Combustible properties of treated wood by fire-tube apparatus. NAHB National Research Center. 1990. Home builders guide to fire retardant treated plywood. Evaluation, testing, ASTM D2898. Accelerated weathering of fire-retardant and replacement. Upper Marlboro, MD: National Association treated wood for fire testing. of Home Builders, National Research Center. ASTM D3201. Hygroscopic properties of fire-retardant wood and wood-base products. Winandy, J.E. 1995. Effects of fire retardant treatments after 18 months of exposure at 150°F (66°C). Res. Note FPL– ASTM D5516. Standard method for evaluating the me- RN–0264. Madison, WI: U.S. Department of Agriculture, chanical properties of fire-retardant treated softwood ply- Forest Service, Forest Products Laboratory. wood exposed to elevated temperatures. ASTM D5664. Standard method for evaluating the effects Winandy, J.E. 1997. Effects of fire retardant retention, of fire-retardant treatments and elevated temperatures on borate buffers, and redrying temperature after treatment on strength properties of fire-retardant treated lumber. thermal-induced degradation. Forest Products Journal. 47(6): 79–86. AWPA. [Current edition]. Gransbury, TX: American Wood- Preservers’ Association. Winandy, J.E.; LeVan, S.L.; Ross, R.J.; Hoffman, S.P.; McIntyre, C.R. 1991. Thermal degradation of fire- Standard A2. Analysis of waterborne preservatives and fire- retardant-treated plywood—development and evaluation of a retardant formulations. test protocol. Res. Pap. FPL–RP–501. Madison, WI: U.S. Standard A3. Determining penetration of preservative and Department of Agriculture, Forest Service, Forest Products fire retardants. Laboratory. Standard A9. Analysis of treated wood and treating solu- Winandy, J.E.; Lebow, P. K.; Nelson, W. 1998. Predict- tions by x-ray spectroscopy. ing bending strength of fire-retardant-treated plywood from Standard C20. Structural lumber—Fire-retardant pressure screw-withdrawal tests. Res. Pap. FPL–RP–568. Madison, treatment by pressure processes. WI: U.S. Department of Agriculture, Forest Service, Forest Standard C27. Plywood—Fire-retardant treatment by Products Laboratory. pressure processes. Standard P17. Fire-retardant formulations.

17–16 From Forest Products Laboratory. 1999. Wood handbook—Wood as an engineering material. Gen. Tech. Rep. FPL–GTR–113. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory. 463 p. IGNITION AND CHARRING TEMPERATURES OF WOOD

By 1 Forest Products Laboratory, Forest Service U. S. Department of Agricultme

-----

A review of the technical literature reveals but a limited amount of data con- cerning minimum temperatures required to produce charring or ignition of wood. Results obtained by different investigators for ignition temperatures show wide discrepancies. As Brown ( 1 )2 has indicated, the different values reported may be due to the specific test conditions associated with the methods employed, and also to the different interpretations among investigators as to what consti- tutes “ignition temperature.” Available published reports ( 1 ), ( 2 ), ( 4 ), ( 5 ), and ( 7 ) on investigations of the ignition of wood usually deal with temperatures, size of material, and rate of air supply in the range that will cause ignition within a few minutes. No available publications relate to long exposures at the lower ranges of elevated temperature to which wood may often be subjected in actual use conditions,

The purpose of this report is to indicate the importance of time in the effects of heat upon wood rather than to present specific values for ignition tempera- tures or to recommend methods for determining such temperatures. A previous investigation by R. E. Prince ( 7 ) demonstrated clearly that what he termed the “ignition temperature” for wood does not have a fixed value but is greatly influenced by the duration of exposure. In that work, oven-dry wood specimens 1-1/4 by 1-1/4 by 4 inches were exposed continuously to different temperatures maintained constant in an electrically heated apparatus. Record was made of the time that the specimens had to be kept at a specified tempera- ture before the gases issuing from the specimens could be ignited by a pilot flame located about one-half inch above the test sample. The results reported for specimens of different species are shown in table 1.

1 Maintained at Madison, Wis., in cooperation with the University of Wisconsin. 2 Numbers in parentheses refer to literature cited at the end of this report.

Report No. 1464 (Rev.) -1- Agriculture-Madison The data are somewhat erratic, especially at the lower temperatures, and they fail to show a consistent relation of ignition time to the specific gravity of the wood. It is probable that further, more comprehensive, testing may remove some of the apparent inconsistencies. Earlier studies, in which exposure was made at gradually increasing temperatures, showed that, as a general rule, a species of low specific gravity could be expected to ignite more readily than one of high specific gravity, provided that the specimens did not vary greatly in their content of resin or other extractive materials that would influence their behavior. It was also shown that ignition might either be hastened by the presence of flammable oils or resins, or be retarded by the presence of other extractives. Aside from these exceptional cases, specific gravity of the test piece was considered more important than species characteristics in influencing ignition when the size of the specimen, moisture content, and conditions of fire exposure were identical.

Some exploratory tests at the Forest Products Laboratory conducted more recently at a lower range of elevated temperature have demonstrated further the importance of time on the behavior of wood heated continuously. In these tests, small kiln- dried, hard maple motor wedges, about 1/8 by 1/4 by 3 inches which were to be used in a special motor for hot-air ducts, were subjected to temperatures rang- ing from 107° to 150° C. for various extended periods in electrically controlled drying ovens. With prolonged exposure at all of the temperatures used there was a gradual darkening of the wood, accompanied by 10SS of weight and shrinkage in the trans- verse dimensions of the specimen. Chemical destruction of the specimens, as in- dicated by their loss of weight, was not associated with any one critical temper- ature. Instead, at each temperature of exposure, the specimens lost weight at a rather regular rate, and the rate became faster as the temperature was raised. Samples which had been exposed to 107° C. for 1,050 days assumed a light choco- late shade. Those exposed to 120° C. for 1,235 days became appreciably embrit- tled, were of a dark chocolate color, and when moistened were strongly acid to litmus paper. Those exposed to 140° and 150° C. had the appearance and friabil- ity of charcoal even before they had lost 65 percent of their original air-dry weight at 6 to 8 percent moisture content, but none was ignited during its ex- posure. A summary of weight losses and transverse shrinkage for different heating periods is given in table 2.

Although comparable data are unavailable, experience leads to the belief that other species would perform in much the same general manner as the maple used in these tests.

The fact that ignition did not occur at any time during this series of tests is no guarantee that it could not have done so if conditions more favorable for com- bustion had prevailed. In the lower range of temperature values, decomposition proceeded so slowly that the gaseous products evolved were dissipated in the surrounding air. In a confined space, however, the opportunity for escape of the

Report No. 1464 -2- gases and the heat accompanying oxidation would be lessened, and the danger of developing spontaneous ignition would be increased. This may account for the fires that have been reported to have started in wood in direct contact with low-pressure steam pipes or in wood heated at temperatures below that where the exothermic reaction normally becomes a factor ( 9 ). There are also indications from experience with wood in dry kilns, steam tunnels, and other places that long continued intermittent heating and exposure to damp conditions accelerate the decomposition of wood. Little detailed information is available on the amounts and composition of the products formed at the temperatures and exposures described in table 2. Klar ( 3 ) reported that upon heating wood between 150° and 200° C., the composition in percent by volume of noncondensable gases is 68 percent carbon dioxide, 30.5 percent carbon monoxide, and 2 percent hydrocarbons. Murphy’s investigation ( 6 ) of the thermal decomposition of paper below ignition temperatures also shows the evolution of gases to be a function of time and temperature.

Report No. 1464 -3- Literature Cited

Report No. 1464 -4- • 5-7 Report No. 1464

GUY E. BURNETTE, JR., P.A. Attorneys at Law 3019 Shannon Lakes North Heritage Oaks Business Center Suite 201 Tallahassee, FL 32309 850/668-7900 850/668-7972 www.gburnette.com

February 11, 2005

CLIENT BULLETIN DAUBERT REVISTED: BITLER v. A.O. SMITH, CORP.

In the twelve (12) years since the U.S. Supreme Court’s landmark decision in Daubert v. Merrill-Dow Pharmaceuticals, courts have directed intense scrutiny at the methodologies and procedures of expert witnesses in all types of cases. The emphasis on “reliability” under the Daubert decision has led to a more restrictive process for judging the admissibility of expert testimony. Although the very premise of Daubert was that the Federal Rules of Evidence were intended to facilitate the admission of expert testimony at trial, this new evidentiary standard has made the admissibility of expert testimony far more difficult than ever before.

A recent decision from the United State Court of Appeals for the Tenth Circuit may signal a new direction for the courts in evaluating expert testimony. In Bitler v. A.O. Smith, Corp., 10th Cir., No. 02-1527 (December 6, 2004) a Daubert challenge was raised in a gas explosion case. Mr. Bitler was severely burned when a gas water heater exploded in his basement. The gas service supply line was an unsupported flexible copper tubing along the ceiling joist in the basement. A T-fitting supplied gas to the water heater and a nearby space heater. In the investigation of the accident, it was revealed that there were leaks in the gas supply line at the T-fitting and the space heater, as well as the safety valve seat of the water heater. A fire investigator hired by an insurance company determined the water heater was the source of the explosion. An investigator hired by the Bitler family determined the explosion was caused by copper sulfide contamination on the safety valve seat of the water heater. Bitler sued the manufacturer of the water heater and the gas company which had installed the water heater and the gas piping.

The Bitlers’ expert identified copper sulfide particles and grease on the safety valve seat which had caused a leak allowing gas to escape. The expert testified that the valve seat had been turned to the “off” position after the explosion and was damaged. It could not be adequately tested afterward to verify the theory of the

1 explosion. Instead, the expert presented his theory through the “elimination method” often employed in fire and explosion investigations to determine the cause of an incident by eliminating all other potential causes. He noted that there had been copper sulfide contamination to the valve seat which would have allowed gas to escape. He eliminated any other source of a gas leak in the immediate area which could have caused the explosion. Using deductive reasoning through the “elimination method”, he concluded the explosion had been caused by the copper sulfide contamination at the valve seat.

Experts hired by the manufacturer and the gas company asserted this theory was not scientifically valid and did not satisfy the reliability criteria under Daubert. The primary point of contention was that the theory could not be validated by testing the valve seat after the explosion incident. Under the Scientific Method outlined in NFPA 921 and embraced by the U.S. Supreme Court in Daubert, a hypothesis or theory must be tested using empirical data in order to validate the theory as reliable.

The trial court excluded the testimony of the expert for the Bitlers based upon this point. On appeal to the 10th Circuit, this decision was overturned. The appellate court ruled that an expert does not have to prove his theory is “undisputedly correct” or universally accepted, despite the language in the Daubert decision requiring scientific verification and employing “general acceptance” as a measure of reliability. Instead, the court held that an expert must only show that his methodology is scientifically sound and his conclusion is based upon facts which can be reasonably established from the evidence. The determination of “relevance and reliability” to admit an expert’s testimony can be made from an examination of the relationship between the methodology, the conclusions and the facts considered by the expert. While the United States Supreme Court in the Kumho v. Carmichael decision had ruled that an expert cannot validate his theory and conclusions based upon the “ipse dixit” of his own opinions, personal experience and professional training will be considered in determining whether a conclusion drawn from deductive reasoning is valid and reliable. The court noted that the concepts of testing and peer review outlined in the Daubert decision may not be appropriate when examining a case where deductive reasoning has been employed. Demonstrating extensive experience, training, proper methodology and deductive reasoning to reach a conclusion will be considered “scientifically valid”. The court observed that testing could not be done with the valve seat because of the damage caused by turning it to the “off” position after the explosion. The inability to conduct an actual test was not fatal to the methodology of the expert because of the unique nature of the explosion incident at issue. The testing of a hypothesis is aimed at “theories which explain causal relations among regularly occurring natural phenomena”. The explosion in this case was a “one-time occurrence” which made testing both unnecessary and inappropriate. The court analogized the process to the use of a “differential diagnosis” in medicine when symptoms suggests several potential causes of a health problem. Considering those potential causes and eliminating them through examination and/or testing is a scientifically sound methodology for arriving at the

2 ultimate diagnosis. In fire and explosion cases, the same approach can be employed as a reliable methodology under the standards of Daubert. The court described the process as “reasoning to the best inference”.

In this case, the expert reliably proved that contamination by copper sulfide particles could cause a leak. He proved there had been contamination to the valve seat. The underlying theory was established to the satisfaction of the court. It then become a question of whether the theoretical potential of a leak had actually occurred in this case. The expert reliably proved through inference and deductive reasoning that this was the only credible explanation for the explosion incident. “Physical investigation, professional experience and technical knowledge” were recognized by the court as the appropriate means for proving the theory.

The Bitler case is significant in several respects. First, it is a decided departure from the strict “testing and validation” requirements which have been imposed in other cases. The expert was allowed to use subjective elements in the investigation of the explosion incident. The court held that the entire process does not have to be purely objective and scientifically provable. Instead, a scientifically sound foundation for the theory with reliable processes used to confirm the theory will be considered as reliable under the Daubert standards.

In sanctioning this process, the court recognized the unique challenges of analyzing and reconstructing a fire or explosion incident. In many of those cases, it will not be possible to recover all of the physical evidence and test it to prove the theory. The Bitler decision will allow an investigator to use a subjective process based upon sound scientific reasoning to reach the ultimate conclusion of the fire or explosion incident’s cause.

At least in the 10th Circuit, expert testimony in fire and explosion cases will be more readily admissible. The case precedent of this decision will likely lead to similar rulings in other circuits across the country. It appears that we may have reached the “high-water mark” of Daubert challenges to the findings of expert witnesses and the tide may have begun to turn. While only time will tell if this truly represents a new trend in the evaluation of expert testimony, it is an encouraging sign that courts will be more receptive to the processes of fire and explosion investigations in future cases.

3 United States Court of Appeals FOR THE EIGHTH CIRCUIT ______

No. 03-3836 ______

Fireman's Fund Insurance * Company; * * Plaintiff, * * Travelers Indemnity Company of * America; * * Intervenor Below/ * Appeal from the United States Appellant, * District Court for the * District of Minnesota. American Economy Insurance * Company; American International * Recovery; Home Video of * Minneapolis, Inc.; * * Intervenors Below, * * v. * * Canon U.S.A., Inc., * * Appellee. *

______

Submitted: October 22, 2004 Filed: January 12, 2005 ______

Before BYE, LAY, and GRUENDER, Circuit Judges. ______GRUENDER, Circuit Judge.

Travelers Indemnity Company of America (“Travelers”) appeals the district court’s1 grant of summary judgment to Canon U.S.A., Inc. (“Canon”) on Travelers’ claims of strict product liability; negligent design, manufacturing and testing; and breach of warranty. For the reasons discussed below, we affirm.

I. BACKGROUND

On October 16, 2000, a fire destroyed Home Video, a video rental store located in a strip mall in St. Paul, Minnesota, and damaged the three other businesses in the strip mall. The mall’s owner was insured by Travelers, the sole appellant in this case. The insurers of the three other tenants, as well as Home Video, were also plaintiffs in the suit below.

A Canon model NP 6016 copier was located in the storeroom of Home Video. The copier had been in use for five years. Service records indicated the copier was upgraded in 1998. Home Video employees stated that the copier often jammed on the heavy paper used for making video cassette jackets but had only jammed once on plain paper in five years. The employees could usually clear paper jams themselves but occasionally had to call a service technician. The copier was serviced one week before the fire.

At around 1:15 p.m. on the day of the fire, a Home Video employee set the copier to make about 80 plain paper copies and left the storeroom with the copying in progress. The locked storeroom apparently was undisturbed until about 7:30 p.m., when another employee entered the storeroom to put some papers in the adjoining

1The Honorable David S. Doty, United States District Judge for the District of Minnesota.

-2- office. The employee did not see, hear or smell anything unusual. The employee then smoked a cigarette. The employee claims to have left the storeroom and entered a back hallway to smoke, although surveillance cameras show him returning to the front of the store from the direction of the storeroom, not from the back hallway. The employee also left the storeroom door locked.

At approximately 8:00 p.m., the store telephones began ringing nonstop, the computers froze, and the security alarms activated. A customer reported the smell of smoke to Home Video employees, and smoke was spotted in the back of the store. The employees opened the door to the storeroom and witnessed the fire in progress. The employees then evacuated the building. The fire destroyed the store and damaged the other stores in the strip mall.

Several fire investigators examined the scene of the fire. The St. Paul Fire Department concluded that the fire was unintentional and that the copier was the most probable cause of the fire. Three other fire scene investigators, hired separately by Travelers, Home Video and two other tenants’ insurers, also identified the copier as the source of the fire. Travelers and the other plaintiffs brought suit against Canon on theories of strict product liability; negligent design, manufacturing and testing; and breach of warranty.

The plaintiff insurance companies hired several fire causation experts to determine how the copier could have caused the fire. The burned copier was subjected to five detailed inspections between early 2001 and September 2002, including visual, x-ray and electron-microscope examinations. Fire causation experts Beth Anderson and Michael Wald each produced reports in October 2002 stating that the copier’s internal burn patterns showed that the upper fixing heater assembly caused the fire and that the design of the assembly was defective because it included a thermal fuse safety device that was not properly rated to prevent such a fire.

-3- Canon’s expert, Lawrence Sacco, filed an expert report challenging the plaintiffs’ theory. In March 2003, Anderson and Wald each filed a rebuttal of Sacco’s opinion in which they introduced the copier’s composite power supply board as another potential cause of the fire. The plaintiffs sought to re-open discovery in March 2003 in order to obtain more information for their composite power supply board theory, but the district court denied the motion as untimely. That decision was not appealed.

Canon moved for summary judgment on the basis that the expert opinions of Anderson and Wald were inadmissible, leaving the plaintiffs with no evidence of a defect, a necessary element of each of Travelers’ claims. The district court granted Canon’s motion, concluding that the expert opinions were unreliable and potentially confusing to a jury. The district court further held that, even if the expert opinions were admitted into evidence, the plaintiffs could not demonstrate that the alleged defects caused the fire. Travelers appeals the district court’s grant of summary judgment to Canon.

II. DISCUSSION

We review the district court’s grant of summary judgment de novo, applying the same standard the district court applied. Anderson v. Raymond Corp., 340 F.3d 520, 524 (8th Cir. 2003). We view the evidence in the light most favorable to Travelers, giving it the benefit of all reasonable inferences that may be drawn from the evidence. Id. We review the district court’s decision concerning the admission of expert opinions for an abuse of discretion. Id. at 523.

A. Reliability of the Expert Testimony

The opinion of a qualified expert witness is admissible if (1) it is based upon sufficient facts or data, (2) it is the product of reliable principles and methods, and (3)

-4- the expert has applied the principles and methods reliably to the facts of the case. Fed. R. Evid. 702. A trial court must be given wide latitude in determining whether an expert’s testimony is reliable. See Kumho Tire Co. v. Carmichael, 526 U.S. 137, 152 (1999).

Anderson and Wald purportedly followed standards set forth by the National Fire Protection Association in its publication NFPA 921: Guide for Fire and Explosion Investigations (1998). This guide qualifies as a reliable method endorsed by a professional organization. See Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 594 (1993). However, NFPA 921 requires that hypotheses of fire origin must be carefully examined against empirical data obtained from fire scene analysis and appropriate testing. The district court did not abuse its discretion in concluding that Anderson and Wald did not apply this standard reliably to the facts of the case.

Anderson and Wald initially stated, to a reasonable degree of engineering certainty, that the burn patterns inside the copier established the copier’s upper fixing heater assembly as the cause of the fire. They attempted to demonstrate that the copier’s safety devices were improperly designed to prevent such a fire.

A brief description of the upper fixing heater assembly is necessary. In normal copier operation, the heating element in the upper fixing heater assembly applies heat to affix the copied image to the paper. The copier’s heater control circuitry varies the electrical current supplied to the heating element to automatically control the amount of heat generated. The heater control circuitry includes several safety features to prevent overheating, including programmed shutdown limits based on feedback from two temperature sensors and independent hardware shutdown limits based on the two sensors. In addition, if the environmental temperature near the heating element persistently exceeds a safety threshold, a thermal fuse in the circuit opens, cutting off

-5- the electrical current to the heating element and stopping any heating. As a final safety measure to prevent overheating, the heating element is designed to self-destruct in the event of rapid heating.

Anderson and Wald each stated that a fire in the upper fixing heater assembly must have started with a malfunction in the heater control circuitry. They concluded that a defective thermal fuse design failed to prevent the fire. This conclusion was based on three experimental tests of an exemplar upper fixing heater assembly in which the heater control circuitry was entirely bypassed, except for the thermal fuse. By carefully applying electrical current directly to the heating element, Anderson was able to produce a thin brown scorch line on a sheet of paper fastened to the heating element before the thermal fuse opened to shut off the current.

We agree with the district court that this experimental testing did not meet the standards of NFPA 921. Anderson and Wald admitted that to actually start a fire without a bypass of the heater control circuitry and its embedded safety features, the heater control circuitry first would have to malfunction. This undescribed malfunction would have to supply an electrical current to the heating element precisely tailored to generate not just scorching, but also an open flame. Furthermore, the temperature rise would have to be fast enough to avoid triggering the thermal fuse yet slow enough to avoid cracking the heating element. See Weisgram v. Marley Co., 169 F.3d 514, 521 (8th Cir. 1999) (excluding an expert’s opinion that a defective thermostat in a baseboard heater caused a fire because the expert could not adequately demonstrate how a backup high-limit control failed), aff'd, 528 U.S. 440 (2000). Not only did the experimental testing fail to produce an open flame, but the experts were unable to explain the assumed heater control circuitry malfunction in theory or replicate it in any test. In short, the experimental testing of the heating element and thermal fuse in isolation did not establish that the thermal fuse would fail to prevent a fire caused by a heater control circuitry malfunction.

-6- Additionally, examination of the thermal fuse in the burned copier revealed that no electrical current was flowing to the heating element when the fuse opened. In other words, the heating element was not activated when the rising environmental temperature caused the fuse to open, suggesting that the heating element was not the source of the fire. NFPA 921 § 2-3.6 requires the investigator to “compare[] his or her hypothesis to all known facts,” but Anderson and Wald did not attempt to reconcile this empirical evidence with their theory.

Travelers also challenges the district court’s finding that the experts’ last- minute alternative theory–a failure of the copier’s composite power supply board– lacked evidentiary support on the record. In normal copier operation, the composite power supply board receives power from a standard electrical outlet via the copier’s power cord. The composite power supply board conditions and distributes the electrical power to the rest of the copier.

After learning of two separate incidents involving Canon copiers of the same model, Anderson and Wald each introduced the composite power supply board theory in their respective rebuttal reports. In one of these incidents, a composite power supply board was observed emitting sparks. The other incident involved an actual fire originating at the composite power supply board, but investigators there noted evidence that the board had been tampered with prior to the fire.

Wald’s rebuttal report stated that the burn patterns inside the copier, combined with his new knowledge of the separate incidents, were “very compelling” evidence that the composite power supply board was the source of the fire. However, Wald did not claim that any particular design or manufacturing defect on the board caused the fire. Furthermore, in his original report, Wald relied on the burn patterns inside the copier to establish the upper fixing heater assembly, located elsewhere in the copier, as the source of the fire. We agree with the district court’s conclusion that this sudden reversal of opinion regarding the meaning of the burn pattern evidence, in a

-7- case where that evidence was the sole basis from which to infer the location of a defect, seriously undermines the reliability of the experts’ opinions.

Anderson similarly changed her opinion of the burn pattern evidence inside the copier after learning of the two other composite power supply board incidents. Her rebuttal report simply stated that more information would be helpful in determining whether the board was involved in the fire, but a motion for additional discovery was denied as untimely. Anderson later tested electrical components that were “of a similar type” to components on the composite power supply board. By applying AC voltage to components designed for DC voltage, Anderson was able to force the components to spark. However, Anderson did not describe any design or manufacturing defect on the composite power supply board that would expose such components to AC voltage. Anderson also admitted that the circumstances surrounding the two other composite power supply board incidents were substantially dissimilar to the events surrounding the Home Video fire.

In summary, neither Anderson nor Wald proposed a specific defect on the composite power supply board that might have caused the fire. Furthermore, neither expert carefully examined this hypothesis of fire origin against empirical data obtained from fire scene analysis and appropriate testing, as required by NFPA 921. The district court did not abuse its discretion in concluding that the evidentiary support for the composite power supply board theory was inadequate.

Because the experts did not apply the principles and methods of NFPA 921 reliably to the facts of the case, the district court did not abuse its discretion in concluding that Anderson’s and Wald’s expert opinions were unreliable. As a result, it was not error to exclude these expert opinions.

-8- B. Potential of the Expert Testimony to Confuse the Jury

In addition to concluding that the expert opinions were unreliable, the district court excluded the opinions on the alternative basis that the underlying experimental testing would be confusing to the jury.

The trial court may exclude evidence if it determines the evidence would confuse the issues or mislead the jury. Fed. R. Evid. 403. The admissibility of experimental tests in product liability cases “rests largely in the discretion of the trial judge and his decision will not be overturned absent a clear showing of an abuse of discretion.” McKnight v. Johnson Controls, Inc., 36 F.3d 1396, 1401 (8th Cir. 1994). “[E]xperimental evidence falls on a spectrum and the foundational standard for its admissibility is determined by whether the evidence is closer to simulating the accident or to demonstrating abstract scientific principles.” Id. at 1402. The more the experiment appears to simulate the accident, the more similar the conditions of the experiment must be to the actual accident conditions. Id.

As described above, Anderson and Wald performed three tests on an exemplar copier and its components. They isolated the heating element and thermal fuse from the upper fixing heater assembly by bypassing the heater control circuitry and providing an electrical current of their choosing to the heating element. In two of these tests, the heating element and thermal fuse were tested outside the copier. The third test was performed with the upper fixing heater assembly physically installed in the exemplar copier, although the heating element and thermal fuse were still operationally isolated from the heater control circuitry and its embedded safety features.

We agree with the district court that these tests appear to recreate the cause of the fire while failing to address the presumed malfunction of the heater control circuitry and its associated safety features. In particular, the experiment on

-9- operationally isolated components while they were physically installed in an exemplar copier could lead a juror to believe that the test results were representative of actual copier operation at the time of the fire. Therefore, the district court did not abuse its discretion in excluding the experts’ opinions on the basis that the tests of the upper fixing heater assembly would be confusing to the jury.

C. Causation

To recover on a claim of strict product liability under Minnesota law, a plaintiff must present evidence from which a jury could justifiably find that (1) the product was in a defective condition, unreasonably dangerous for its intended use, (2) the defect existed when the product left defendant’s control, and (3) the defect was the proximate cause of the injury sustained. Lee v. Crookston Coca-Cola Bottling Co., 188 N.W.2d 426, 432 (Minn. 1971). Claims of negligence and breach of warranty also include the causation element. See Myers v. Hearth Techs., Inc., 621 N.W.2d 787, 792 (Minn. Ct. App. 2001) (setting forth the elements of a negligence claim); Alley Constr. Co. v. State, 219 N.W.2d 922, 925 n.1 (Minn. 1974) (setting forth the elements of a breach of warranty claim).

Absent the excluded opinions of the fire causation experts, Travelers presented no evidence of any defect in the copier. Because Travelers cannot prove that a defect in the copier was the proximate cause of the fire, summary judgment for Canon was proper. See Lee, 188 N.W.2d at 432 (“[T]he mere fact of injury during use of the product usually is insufficient proof to show existence of a defect at the time defendant relinquished control.”).

Furthermore, even if the expert opinions of Anderson and Wald had been admissible, we agree with the district court that Travelers failed to present evidence from which a reasonable jury could find that a defect in the copier was the proximate cause of the fire. The experts theorized that the thermal fuse was defective.

-10- However, the experts’ experimental tests did not demonstrate that the heating element could generate an open flame before the thermal fuse opened, and the experts admitted that an open flame would have been necessary to start the fire. Therefore, the experts failed to demonstrate that the thermal fuse was defective.

In addition, the experts admitted that the heater control circuitry would have to malfunction in order to supply enough electrical current to the heating element to start a fire. However, the experts advanced no theory or experiment showing how the heater control circuitry could malfunction to produce such a current. Without evidence to show that the heater control circuitry could malfunction in such a way as to start a fire, Travelers cannot show that a defectively designed thermal fuse failed to prevent that fire. Therefore, even if the expert opinions had been admissible, we conclude that Travelers produced no evidence from which a reasonable jury could find that the allegedly defective thermal fuse caused the fire.

Similarly, neither Anderson nor Wald advanced a theory or experiment showing how a defect on the composite power supply board could have caused the fire. Therefore, we also conclude that, had the expert opinions been admissible, Travelers produced no evidence from which a reasonable jury could find that the allegedly defective composite power supply board caused the fire.

Travelers claims that, had Canon employed an alternative design for the copier incorporating additional thermal and electrical fuse protection, a fire could never have started. Travelers relies on Lauzon v. Senco Prods., Inc., 270 F.3d 681 (8th Cir. 2001), for the proposition that if a plaintiff identifies an alternative design that would have prevented the accident, then the design of the product involved in the accident must be a proximate cause.

In Lauzon, the manufacturer produced two models of a nail gun. The “SN2” model allowed the operator to rapid-fire nails by bouncing a contact point on the nose

-11- of the gun against the work surface while squeezing the trigger continuously. The sequential-fire model, on the other hand, required the operator to release and depress both the trigger and the nose contact point each time to fire a nail. Lauzon drove a nail through his hand when his SN2 recoiled during rapid-fire mode. An expert tested the SN2 and determined that such an accident would not have been possible using the sequential-fire model.

The district court held the expert’s testimony inadmissible, in part because the expert was unable to rule out other accident theories. Id. at 693. This Court reversed, holding that the expert “ruled out all other possible explanations through a safer alternative design, the sequential-fire pneumatic nailer.” Id. The expert’s testing established that the defective design of the SN2 caused the accident because the “use of the sequential-fire tool would preclude a nail being expelled at all, let alone into the hand of Lauzon.” Id. at 694.

The instant case is distinguishable from Lauzon on two important grounds. First, the experimental testing in Lauzon proved that the rapid-fire mechanism was a but-for cause of the accident. In contrast, the experimental testing performed by Anderson and Wald produced no evidence that inadequate fuse protection in the copier was a but-for cause of the Home Video fire. Second, in Lauzon the safer alternative design was embodied in an existing product by the same manufacturer, and the expert showed specifically how its design would have prevented the accident. In the instant case, the experts talked vaguely about adding more fuses, but offered no evidence of a workable alternative design that would have made a fire less likely. Therefore, Lauzon does not help Travelers establish the causation element of its claims.

In summary, with or without the inadmissible opinions of its fire causation experts, Travelers did not introduce evidence from which a reasonable jury could find

-12- that a defect in the copier was the proximate cause of the fire. Therefore, summary judgment for Canon was proper.

III. CONCLUSION

We affirm the district court’s grant of summary judgment to Canon on Travelers’ claims of strict product liability; negligent design, manufacturing and testing; and breach of warranty. ______

-13-