interFIRE Home interFIRE Home interFIRE VR Support Training Calendar Training Center Resource Center Message Board Insurance Info

Glass Breakage in Fires

Dr. Vytenis Babrauskas
April 28 1997.
Copyright 1997
Fire Science and Technology Inc.


The size (more technically, the heat release rate) of fires is limited by the flow of oxygen available to it. In all except very rare circumstances, the flow of oxygen into a room comes largely from open doors and open windows, and to a slight extent from any mechanical ventilation systems and from building leakage. Once a fire gets going, however, windows previously closed may crack and break out. Or...they may not. The results will often be drastically different, depending on whether the windows break or not. Thus, it becomes of significant interest to be able to predict if, and when, glass may break out.

Here, an important distinction needs to be made. When a window pane of ordinary float glass is first heated, it tends to crack when the glass reaches a temperature of about 150 - 200C. The first crack initiates from one of the edges. At that point, there is a crack running through the pane of glass, but there is no effect on the ventilation available to the fire. For the air flows to be affected, the glass must not only crack, but a large piece or pieces must fall out.

Understanding the conditions under which pieces actually fall out has been of considerable interest to many persons concerned with fire. Since the fire ventilation openings need to be known in order for fire models to be used, glass breakage has been of special interest to fire modelers. This has prompted a number of theoretical and simplified studies and a few empirical ones as well.

It must be realized that there are at least two distinct types of thermal exposure to glass that is involved in fires:

  1. A window is inside a room in which a fire is taking place. The window is being subjected to ijavascript:MMersion heating from one side. The local gas temperature and the radiating temperature are rather similar. There may be a gradient of temperature and heat flux from the top down to the bottom.
  2. A window is exposed to an outside fire, typically a wildland or bush fire. In that case, there may be relatively little difference in exposure between the top and the bottom of the window. The heating is primarily by radiation. Local gas temperatures may be near-ambient, since flames are not directly washing against the window and there is a convective cooling flow along the surface.

Theoretical and experimental studies of glass cracking in fires

Keski-Rahkonen [1] presented the first extensive theoretical analysis of glass cracking in fires. He identified that temperature differences between the exposed glass surface and the glass shielded by the edge mounting play the dominant role in controlling cracking. His theory predicted that a temperature difference of about 80C between the heated glass temperature and the edge temperature is needed to initiate cracking. Pagni and Joshi [2] extended these ideas to include more heating physics and an expanded consideration of glass thermal properties. They predicted 58C as the temperature difference for crack initiation. The difference was largely due to assuming different thermal and mechanical properties for glass. Skelly [3] conducted a series of experiments in an unusual small-scale fire test room. One peculiarity of his tests was that the windows were never exposed to a vertical temperature gradient. He reported some fall-out of glass sections, but did not provide any guidelines or tabulations to assist in determining conditions leading to glass breaking out.

Mowrer [11] presented the latest experimental study on the subject, heating both large-scale and small-scale specimens with a radiant panel to simulate a wildland, external exposure. The maximum heat fluxes, which went up to 16 kW m-2, were sufficient to cause cracking, but not breaking out of window panes. Cracking of single-strength glass was found to occur at 4 to 5 kW m-2. Either black or bright insect screens raised by about 21% the heat flux needed to cause cracking. He also found that approximately 33% of the radiation incident on a single-strength pane of glass is transmitted through it. This information can be of use in estimating ignitions inside a building from external radiation.

The National Research Council of Canada (NRCC) has had a program for developing sprinkler protection for glazing. As part of that work, a few non-sprinklered tests have been run where glass 6 javascript:MM thick tempered glass was exposed to simulated room fire conditions, but without a sprinkler [9]. While such glass type would only be cojavascript:MMon in cojavascript:MMercial buildings, the results are nonetheless of interest. Tempered glass behaves differently, in that it shatters upon initial cracking, but the initial cracking does not occur until the glass reaches rather high temperatures. An exposed-surface temperature of 290-380C has been found to be needed, with the unexposed surface temperatures being about 100C lower. Such glass temperatures are normally not reached until after room flashover has occurred. In a later study [10], NRCC examined glazing using radiant heat exposure. Under such conditions, "plain" glass of unspecified thickness was found to "break" when the exposed side reached 150-175C, with the unexposed side being at 75-150C.

Experiments and guidance on glass breaking out in fires

The earliest guidance to be found in the literature on the question of when glass breaks out in fires comes from the Russian researcher Roytman [4] who notes that a room gas temperature of around 300C is needed to lead to glass breakage. The research base for this conclusion is unclear, however.

Hassani, Shields, and Silcock [5] conducted a series of experiments in a half-scale fire test room using 0.9 x 1.6 m single-glazed windows where they created a natural top-to-bottom temperature gradient in the room and in the glass. At the time the first crack occurred in 4 or 6 javascript:MM thick glass panes, gas temperatures in the upper layer of 323 - 467C were recorded. By the end of their 20 min tests, gas temperatures were at ca. 500C. Yet in only 1 of 6 tests was there any fall-out of glass. Temperature differences between the glass exposed surface and the shielded portion ranged between 125C to 146C at the time of crack initiation. These temperatures were about twice that predicted from the no-vertical-gradient theories. The authors do not give the exact room fire temperature at which the glass fall-out began in the one test where this occurred, but this had be higher than 431C (crack initiation) and lower than ca. 450C (end of test). One can put these data together, then, to conclude that at a room gas temperature of around 450C the probability is 1/6 for glass to break out. The same authors [15] conducted further tests using a room with two windows glazed with 6 javascript:MM thick panes. In two tests, fire temperatures of 400 - 500C in the vicinity of the glass resulted in no fall-out of glass. In the third test, one pane fell out at a gas temperature of 500C. The gas temperature when the second window's pane fell out was also about 500C, but some 8.5 more minutes had gone by, the temperature being relatively steady in that time.

The newest experimental results concerning glass exposed to a uniform hot temperature come from the Building Research Institute (BRI) of Japan [6]. In that study, researchers used a large-scale high-temperature door-leakage testing apparatus that resembles a large muffle furnace. Only single-glazed, 3 javascript:MM thick window glass was studied. For this type of glass, however, enough tests were run so that a probability graph could be plotted. These researchers' results are presented in terms of a probability of glass breaking out, as a function of temperature rise above ambient. The figure below shows the results.

The Gaussian fit that can correlate this data corresponds to a mean temperature rise of 340C, and a standard deviation of 50C. Thus, the BRI results are somewhere in between the two earlier values.

The Loss Prevention Council of the UK [12] studied room fires which were providing fire exposure to a multi-story faade test rig. Double-glazed windows were examined, with each pane being 6 javascript:MM thick. Using 3 MW wood crib fires, it was found that temperatures of at least 600C had to be sustained for 8 - 10 min before glass started falling out sufficiently so that fire venting would occur. When tests were repeated using a fully-furnished office room arrangement, however, glass broke out at 5 min after the start of fire. In that test, the temperature was also about 600C at the time of failure, but occurred ijavascript:MMediately as the temperature was reached. Thus, the findings lead to the conclusion that double-glazed windows using 6 javascript:MM thick glass will fail at ca. 600C and that, if the fuel load is significant, the failure may be expected to occur essentially at the instant that 600C is first reached.

Shields, Silcock and Hassani [13] exposed two sizes of double-glazed windows to room fires. The glass thickness was 6 javascript:MM. The room fire reached a peak of 750C and no glazing fell out up to the peak. However, during the decay part of the fire, in one of 3 tests with the larger-size window (0.8 x 1.0 m) fall-out of the inner pane occurred at 21 min, when the temperature had dropped to 500C. Glass did not ever fall out from the outer pane, nor did any fall-out occur in the smaller (0.8 x 0.5 m) window, nor did any fall-out occur in the other two tests. The same authors later [14] tested a room having a wall with twelve1.5 x 1.5 m double-glazed windows. The test record is very brief, but it is indicated that total failure of the first window occurred when the gas temperature was at 350C; it is not clear what the temperatures were for the fall-out of the subsequent windows.

For radiant exposure, Cohen and Wilson [7] reported on an interesting series of experiments simulating flame exposure from a wildland fire. They examined small (0.61 x 0.61 m) and large (0.91 x 1.5 m) panes, single- and double-glazed. They also repeated the tests with tempered glass and with double-glazed windows. For the small windows, at their lowest heat flux, 9.3 kW m­2, all windows cracked, but no glass fell out. Even at the highest flux of 17.7 kW m­2 panes did not fall out. For the larger size panes, at fluxes of 16.2 to 50.3 kW m­2, at least one out of 3 test specimens exhibited fall-out. Tempered glass, by contrast, showed no cracking at tests up to 29.2 kW m­2 in the larger size. The authors also did tests on double-glazed windows, which showed better performance. In experiments with large-size double-glazed windows (non-tempered), they found that fluxes between 20 and 30 kW m­2 were required to cause fall-out in both panes.

Harada et al. [17] tested 3 javascript:MM thick float glass (specimen size: 0.5 by 0.5 m) by subjecting them to various heat fluxes from a test furnace. Below 8 kW m-2, no significant fallout occurred, but for a heat flux of 9 kW m-2, in some cases 8 - 24% of the specimen area fell out. Edge constraint did not affect the results.

Additional data are available from the NRCC study [10], where heat-strengthened and tempered glass (unspecified thickness) was found not to break at an irradiance of 43 kW m-2. The latter heat flux corresponded to 350C on the exposed face and 300C on the unexposed face. Thus, this appears to extend Cohen's data point of 29.2 kW m-2 for non-breakage to 43 kW m-2, without actually determining the point at which breakage and fall-out do occur.

Other types of glass

The published studies have dealt primarily with thin panes of annealed or tempered soda glass. Yet, there are many other types of glass to consider. Some very thick plate glass is used in many cojavascript:MMercial buildings. Plate glass of 6 javascript:MM thickness was found to shatter after a significant time (7 min) of exposure to a radiant heat flux of 23 kW m-2 [15]; otherwise information is not available on its performance. Automotive glass is another category which has not been studied in a systematic way. Finally, there are various fire-resistive glasses. The oldest category of the latter is wire glass. Nowadays, several types of patented fire-resistive glasses also exist which are not wired glass. These are usually multi-layered structures, generally involving some polymeric inner layers. Fire-resistive glasses will normally be accompanied by a laboratory report of the endurance period. Such glasses can be assumed to have no ventilation flow until after their failure time.

Effect of window frame type

Actual fall-out of glass from windows is also influenced by the window-frame material. Mowrer [11] found that vinyl-frame windows tended to show a failure of the frame (e.g., the whole assembly collapsing) before significant fall-out of glazing. Vinyl frame failures were observed when heat fluxes came up to the range of 8 to 16 kW m-2. By contrast, McArthur [16] found that glass in aluminum-framed windows showed a tendency to survive longer than did glass in conventional wood-frame windows.


A theory exists for predicting the occurrence of the first crack in glazing, but this is not directly relevant to fire ventilation. The above brief review of the literature shows that it is, in fact, very difficult to predict when glass will actually break enough to fall out in a real fire. The Russian recojavascript:MMendation of 300C appears to be a reasonable lower bound. The new Japanese study can be taken to imply that 3 javascript:MM window glass will break around 340C. For thicker, 4-6 javascript:MM glass, the mean temperature of breakage would appear to be in excess of 450C, although the difference between the thinner and the thicker glass results seems rather larger than one would surmise. Double-glazed windows using 6 javascript:MM glass can be expected to break out at about 600C. Tempered-glass in not likely to break out until after room flashover has been reached.

In terms of external fires, at a heat flux of 9 kW m­2 some experimental results on ordinary glass showed the possibility of significant fallout. Other studies found that higher heat fluxes were needed, thus 9 kW m-2 appears to be a conservative, but realistic lower bound. Double-glazed windows can resist approximately 25 kW m­2 without fall-out. Tempered glass is able to resist fluxes of 43 kW m­2, at least under some conditions.

Factors such as window size, frame type, glass thickness, glass defects, and vertical temperature gradient may all be expected to have an effect on glass fall-out. Over-pressure due to gas explosions is an obvious glass failure mechanism. Yet, normal fires do show pressure variations and these could potentially affect the failure of glass panes. All of these factors deserve some more study to obtain useful, quantitative guidance.

The above review has dealt only with the role of glass breakage in fire ventilation. A number of other aspects of glass breakage are important to fire investigators; these have been well presented by Schudel [8].


[1] Keski-Rahkonen, O., Breaking of Window Glass Close to Fire, Fire and Materials 12, 61-69 (1988).

[2] Pagni, P. J., and Joshi, A. A., Glass Breaking in Fires, pp. 791-802 in Fire Safety Science-Proc. Third Intl. Symp., Elsevier Applied Science, London (1991).

[3] Skelly, M. J., Roby, R. J., and Beyler, C. L., An Experimental Investigation of Glass Breakage in Compartment Fires, J. Fire Protection Engineering 3, 25-34 (1991).

[4] Roytman, M. Ya., Principles of Fire Safety Standards for Building Construction, Construction Literature Publishing House, Moscow (1969). English translation (TT 71-580002) from National Technical Information Service (1975).

[5] Hassani, S. K., Shields, T. J., and Silcock, G. W., An Experimental Investigation into the Behaviour of Glazing in Enclosure Fire, J. Applied Fire Science 4, 303-323 (1994/5).

[6] Tanaka, T., et al., Performance-Based Fire Safety Design of a High-rise Office Building, to be published (1998).

[7] Cohen, J. D., and Wilson, P., Current Results from Structure Ignition Assessment Model (SIAM) Research, presented in Fire Management in the Wildland/Urban Interface: Sharing Solutions, Kananaskis, Alberta, Canada (2-5 October 1994).

[8] Schudel, D., Glass Fracture Analysis for Fire Investigators, Fire and Arson Investigator 46, 28-35 (March 1996).

[9] Richardson, J. K., and Oleszkiewicz, I., Fire Tests on Window Assemblies Protected by Automatic Sprinklers, Fire Technology 23, 115-132 (1987).

[10] Kim, A. K., and Lougheed, G. D., The Protection of Glazing Systems with Dedicated Sprinklers, J. Fire Protection Engineering 2, 49-59 (1990).

[11] Mowrer, F. W., Window Breakage Induced by Exterior Fires, pp. 404-415 in Proc. 2nd Intl. Conf. on Fire Research and Engineering, Society of Fire Protection Engineers, Bethesda, MD (1998). Also: Mowrer, F. W., Window Breakage Induced by Exterior Fires (NIST-GCR-98-751), Natl. Inst. Stand. and Technol., Gaithersburg MD (1998).

[12] Fire Spread in Multi-Storey Buildings with Glazed Curtain Wall Facades (LPR 11: 1999), Loss Prevention Council, Borehamwood, England (1999).

[13] Shields, T. J., Silcock, G. W. H., and Hassani, S. K. S., The Behavior of Double Glazing in an Enclosure Fire, J. Applied Fire Science 7, 267-286 (1997/98).

[14] Shields, T. J., Silcock, G. W. H., and Hassani, S. K. S., The Behavior of Glazing in a Large Simulated Office Block in a Multi-Story Building, J. Applied Fire Science 7, 333-352 (1997/98).

[14] Shields, T. J., Silcock, G. W. H., and Hassani, S. K. S., The Behavior of Single Glazing in an Enclosure Fire, J. Applied Fire Science 7, 145-163 (1997/98).

[15] Moulen, A. W., and Grubits, W. J., Water-Curtains to Shield Glass from Radiant Heat from Building Fires (Technical Record TR 44/153/422), Experimental Building Station, Dept. of Housing and Construction, North Ryde, Australia (1975).
[16] McArthur, N. A., The Performance of Aluminum Building Products in Bushfires, Fire and Materials 15, 117-125 (1991).

[17] Harada, K., Enomoto, A., Uede, K., and Wakamatsu, T., An Experimental Study on Glass Cracking and Fallout by Radiant Heat Exposure, pp. 1063-1074 in Fire Safety Science--Proc. 6th Intl. Symp., Intl. Assn. for Fire Safety Science (2000).

Home | interFIRE VR Support | Training Calendar | Training Center | Resource Center | Message Board | Insurance Info
Sponsorship Opportunities
Web Site Designed for 800 x 600 by Stonehouse Media Incorporated® Copyright © 2022 All Rights Reserved.