Structural Glass Systems under Fire: Overview of Design Issues, Experimental Research, and Developments

Architectural design concepts incorporating glass beams, panels, or generally load-carrying elements and stiffeners for buildings, claddings, windows, and partitions are largely considered in modern high-rise constructions. A multitude of aspects, including motivations related to transparency, aesthetics, illumination, and energy conservation, progressively increased the use and interest for such a still rather innovative constructional material. However, compared to other traditional materials for buildings, standard glass is typically characterized by brittle behaviour and limited tensile resistance. The intrinsic properties of glass, moreover, together with typically limited thickness-to-size ratios for glazing elements, or the mutual interaction of glass components with adjacent constructional elements as a part of full assemblies they belong (i.e., fixing systems, sealants, etc.), as well as the combination of mechanical and thermal phenomena, make glass structures highly vulnerable. Special safety design rules are hence required, especially under extreme loading conditions. In this review paper, a state of the art on structural glass systems exposed to fire is presented. Careful consideration is paid for actual design methods and general regulations, as well as for existing research outcomes—both at the material and assembly levels—giving evidence of current challenges, issues, and developments.


Introduction
Glass is largely used in buildings as a construction material, to replace and/or interact with traditional structural elements composed of steel, aluminum, timber, and concrete. Major applications of glass in buildings are related to a multitude of aspects, including aesthetics, lightening, transparency, and insulation motivations (see, e.g., Figures 1(a) and 1(b)).
Generally, glass is known to behave as a brittle material with relatively high compressive resistance and limited tensile strength, hence shattering into many dangerous shards [1,2]. Fail-safe design concepts, in this sense, are mandatory, both under ordinary loads and extreme loading conditions.
In this regard, several research studies have been dedicated over the last years to the development and/or assessment of speci c design regulations and novel design concepts for structural glass systems, including extended experimental and nite element (FE) numerical investigations related to connections, composite assemblies, and hybrid systems [3][4][5][6].
Especially in the case of re accidents, special enhanced safety levels should in fact be ensured, to allow evacuation of buildings (Figures 1(c) and 1(d)).
Multiple aspects are however combined in the overall re performance of a given structural glass system, such as the typical brittle behaviour of material, the high sensitivity of its mechanical properties to temperature, the high sensitivity of re performance to geometrical features, glass type, and the mutual interactions between all the system components (i.e., the structural glass assembly, including supports and building components).
As a speci c issue of glass systems, in addition, their re performance cannot be analytically derived but requires re testing estimations. Advanced FE modelling, in this regard, could represent a valid alternative to time-and costconsuming experiments. However, major issues for the FE analysis of structural glass elements under re derive from the current lack of standardized guidelines and general rules able to o er reliable results [21], as well as of wellestablished mechanical and thermal properties of materials in use. In addition, FE literature e orts related to the re performance of structural glass systems are very limited (see, e.g., [22]).
In this paper, a review of experimental research on structural glass systems under re is proposed. Section 2 rst recalls a short overview of structural design concepts and requirements. In Section 3, mechanical and thermal properties of standard glass under high temperatures are reported, giving evidence of several literature source outcomes, as well as relatively recent re-resistant (FR) glass solutions available on the market for special applications. Careful consideration is paid, in particular, for material properties representing key input parameters for structural design purposes. Sections 4 and 5 nally present a summary of existing experimental research related to the re performance of glass to high temperatures, including material properties (Section 4) and structural glass systems (Section 5), such as oors and overheads, beams, facades and windows, and glazing systems retro tted via protective lms.

Structural Glass Systems under Ordinary Loads.
Compared to traditional materials in use of constructions, one of the major factors a ecting the design and veri cation of structural glass elements is represented by its intrinsic features. Even without fault of the designer, a given structural glass element can in fact break unexpectedly, during its service life [1]. Whatever the reason, the structural integrity of the overall assembly it belongs must not be compromised. According to the general design concept of EN 1990:2002 [23], both ultimate limit state (ULS) and service limit state (SLS) should be properly veri ed. e ULS resistance veri cation is intended to ful ll the structural safety of a structural glass element [1,2]. Such a safety assessment is generally performed by limiting the maximum principal stresses achieved under relevant load combinations to do not exceed the design resistance of glass. Multiple aspects can a ect, however, the design value of glass resistance (see, e.g., [1,2]), being de ned as a function of glass type, loading (i.e., in-plane and out-ofplane), loading time (i.e., instantaneous, permanent, etc.), edge e ects and treatments, glass surface treatments, pro le, etc. In Europe, following the recommendations of [2,24], several national codes adopted the same design provisions (see, e.g., [25][26][27]). Di erent approaches can be found in US regulations, while further issues also derive from a combination of e ects due to multiple design actions (see, e.g., [28]). e SLS veri cation is aimed at the limitation of deections. e reference limit values for such deformations mainly depend on the speci c applications or support conditions. As in the case of speci cations for ULS design, di erent SLS limit deformation values can be found in standards. A further design condition that should be properly veri ed (see, e.g., the CNR guidelines [29]) is then associated to the so-called collapse limit state (CLS). Given a structural glass system to verify, in order to ensure appropriate redundancy in the case of accidental cracking, the residual CLS resistance and maximum deformations of the partially damaged system are also required.

Structural
Glass Systems under Fire Loading. Fire loading represents, for glass systems as well as for constructions in general, an extreme loading con guration. As such, speci c provisions should be taken into account to ensure appropriate performance levels. Given a glazing system exposed to re, in accordance with existing standards (see, e.g., the EN 13501-2 regulations [30]), its re performance is generally de ned on the basis of three classi cations levels: (a) integrity (classi cation "E"): glass prevents ames, smoke, and hot gases from passing through. e re remains contained; (b) limiting radiation ("EW"): glass restricts the amount of heat passing through it to the side which is to be protected; (c) thermal insulation ("EI"): the average temperature of glass on the protected side remains below 140°C; hence, the risk of self-combustion of exposed materials (due to either radiation or convection) can be minimized, and buildings can be evacuated safely and calmly.
e above FR reference criteria can only be determined on the basis of re experiments, and typical FR rating classes are associated to 30, 60, or 120 minutes of performance. Relevant standards in use in Europe are, for example, the EN 1363-1 document [31], providing FR test requirements and methods; EN 1364-1 [32], for non-load-bearing elements and walls; and EN 1634 [33], for doors and shutters. Floors and roofs should be indeed tested in accordance with EN 1365-2 regulations [34] and then classi ed by following the EN 13501-2 provisions.
In addition to EU provisions, the American Underwriters Laboratory standard [35] includes a further requirement; that is, a given FR glazing system should have the ability to withstand the so-called "hose-stream test," which assesses the system ability to remain intact after a jet of water is blasted on its surface, when exposed to re [36].
Compared to other extreme loading conditions which may occur over the lifetime of a given structural glazing system, the main issue of FR glass systems arises from glass response to temperature variations. Conventional glass for application in buildings (Section 3), in fact, o ers typically limited resistance when exposed to re and generally shatters within minutes, giving evidence of the so-called thermal breakage phenomena (Section 4). Heat treatment can o er slightly longer resistance, but this enhancement could not be signi cantly enough. Special glass types, conventionally detected as "FR glasses" or " re-rated glasses," are indeed available on the market for speci c applications (Section 3.3). On one side, besides the general re performance of standard glasses, past experimental research proved that ordinary glass systems can also o er interesting re performances (Section 5). However, pure thermal e ects combined with additional mechanical loads acting on a given structural glass system to verify should severely compromise its overall performance, hence requiring experimental testing and detailed investigations at the material level as well as at the component and assembly levels. ere, nominal values are also provided for characteristic resistances in tension and compression. As known, several SLS glass types are in fact commercially available [1,2], with annealed (AN) oat glass representing the reference base material. e strength of AN glass is typically limited, compared to other constructional materials, with a nominal characteristic value in tension up to 45 MPa. e mechanical properties of AN glass can then be enhanced via thermal or chemical processes, leading to strengthened (HS, with 70 MPa the nominal tensile resistance value) or fully tempered (FT, with 120 MPa its tensile resistance) glasses, respectively, with improved tensile strength as well as bene cial e ects especially in terms of the shape and size of shards, in the case of accidental failure, due to the initial state of residual stresses resulting from strengthening processes. For the sake of clarity, Table 1 mentions nominal values only of material mechanical properties.

Glass in Constructions
As far as the transition temperature T g is not exceeded, glass behaves linear elastically under the assigned design loads. Given a combination of ordinary loads to verify, as a result, the knowledge of elastic mechanical properties and resistance values for SLS glass given in Table 1 allows then to perform analytical or FE structural analyses.
Crack occurrence and propagation, however, may even prematurely occur due to possible thermal stresses, hence requiring a typical multidisciplinary approach for such a constructional material. ermal shocks, that is, cracking due to temperature gradient between heated and unheated glass regions, as well as due to relatively low thermal expansion coe cient of glass, are in fact typically expected to occur when the temperature gradient lies in the order of 40°C for AN glass, up to 100°C for HS glass, and 200-250°C for FT glass [1]. A huge number of research studies have been focused on thermal failure assessment of glazing windows, taking care of simple glass panels, double glass units, and point-xed systems (Section 4).

Laminated Safety Glass and Insulated Glass.
Further design issues and complexities arise as far as glass systems do not consist of single panes but are assembled in composite laminated sections and/or insulated glass units, as conventionally in use for buildings.
Laminated glass (LG) represents, in general terms, the combination of two or more glass plies together with foils consisting in a certain interlayer type. LG has been rst developed for automotive applications, since early 1900, to avoid injuries in case of accidents, and only in the last decades, LG has been largely used in civil engineering applications for structural purposes. As a general rule of the LG concept, the resisting cross section is expected to respond as a composite system to external loads, hence having enhanced mechanical performances than single glass panes, both in the elastic stage and in the postcracked phase. From a mechanical point of view, the rst implicit advantage of LG structural applications is that multiple glass layers can be bonded together; hence, the required level of resistance, sti ness, and redundancy can be obtained by using conventional glass thicknesses available on the market. In addition, thanks to the presence of bonding lms, LG represents since decades the conventional safety glass solution in buildings, since able to hold together glass shards in case of failure, hence reducing possible risks for people (Figures 2(a) and 2(b)).
Bonding lms typically consist of polyvinyl butiral (PVB) lms, ionoplast foils (i.e., SentryGlas ® (SG)), and ethylene-vinyl acetate (EVA) compounds. As a common aspect of such possible interlayers, besides their di erent constitutive laws, these lms are generally characterized by viscous behaviour; hence, they are generally sensitive to temperature and load-time application, as also emphasized by several research e orts [37][38][39]. In addition, even at room temperatures, interlayers in use for LG applications are generally characterized by relatively low shear sti ness, compared to glass (Figure 2(c)). e overall structural performance of a given LG composite section is hence highly dependent on the interlayer features, including durability and resistance.
In terms of structural design of LG systems under ordinary loads, various methods are available to account for the e ects of interlayer degradation over time and temperature increase (see [1,2] for a summary of existing formulations). As a result, at the design stage, optimal resistance and sti ness behaviours for ULS and SLS are generally ensured, as well as appropriate safety performances for the CLS postcracked stage. Mostly, null contribution, given the typical mechanical degradation of interlayers for temperatures higher than 30°C (Figure 2(c)), is however expected from interlayers under re loading; that is, the LG section is expected to behave fully uncoupled. ere, speci c design assumptions should be taken into account, including the use of special intumescent compounds (Section 3.3).
Multiple glass panes (monolithic and LG sections) can then be assembled together to act as insulated glass units, both double (i.e., single gas cavity interposed between glass panels) or triple (i.e., double cavity). ere, design calculations should take into account the so-called load-sharing e ects due to air or gas in ll in the cavities, that is, the mutual interaction between glass panes once subjected to mechanical loads [1,2]. Ordinary climatic loads represent an additional design condition to properly verify, due to possible variations in the cavity volume and pressure. ermal analyses of insulated glass systems are hence required even under solar exposure only. All the mentioned aspects and variables, consequently, further increase the design complexity for glazing systems under re.

Fire-Resistant
Glazing. FR glazing represents a relatively recent solution, known to provide excellent protection for lives and property in the event of re, and hence may be used as a barrier for re separation or compartmentation (for a speci ed duration), enabling occupants to assemble in a relatively safe compartment and acting as a part of an integrated " re safety strategy" for the full building it belongs. As a crucial aspect of such solutions, FR glass systems require extreme attention in installation Depending on the treatment of glass; b theoretical value, in accordance with [29].
detailing. In addition, all the FR components, such as the glazing seals, beads, xings, and frame, must be compatible and work together to achieve the required performance [36]. FR glazing, due also to relatively high costs, actually nds limited applications in buildings, especially where protected escapes must be ensured in the case of re accidents. In accordance with a study carried out by Yang et al. [40], for example, FR glazing was representing in 2011 less than 5% the overall China glazing applications. Major limitations in the use of FR glass derive also from current need for additional research e orts and investigation on its actual re performance (Section 4).
In accordance with design guidelines such as [36], FR glazing solutions actually available on the market can include (i) LG composites, obtained by bonding together di erent glass types (e.g., SLS glass panes with enhanced and FR glass types) with special re interlayers (i.e., intumescent lms); (ii) wired glass; (iii) ceramic glass; (iv) resin laminated glass; (v) gel laminated glass; and (vi) thermally toughened alkaline earth silicate safety glass ( Figure 3). Multiple glass layers according to (i)-(vi) example types can then be combined in insulated FR glazing units. Within the given list of (i)-(vi) solutions, wired glass elements do not o er enhanced re resistance compared to ordinary glass and typically crack early due to thermal stresses. FR performance is indeed ensured by their integral wire mesh, able to hold together and in place cracked glass pieces.
Generally speaking, for structural design purposes and mechanical calculations, a given FR system can be conventionally treated as a standard, non-FR glass element (Sections 2.1, 3.1, and 3.2), with the di erence of enhanced performance under re exposure. In general terms, FR glasses are in fact considered as e ective passive re protection (PFP) tools for buildings, with speci c applications like glazed internal and external re doors (vision panels); interior partitions and compartments; roofs, oors, and ceilings; façade panels; escape and access corridor walls; and stairways, lobbies, and enclosures (to protect shafts). Juxtaposed with PFP tools, active re protection (AFP) systems can provide further e ort in combination with PFP systems but require a certain motion and response to combat re. Typical AFP tools can be either automatic (i.e., water sprinklers, re alarms, hypoxic air suppression systems, etc.) or manual (i.e., emergency evacuation, re extinguishers, re ghters, water hoses, etc.). Fire design and optimization of such systems, however, is generally complex and requires speci c competences. Combined PFP and AFP systems, moreover, are generally expected to provide enhanced bene ts but could also lead to worst performances. So far, several research studies [41][42][43][44][45] highlighted, for example, that water lms and sprinklers can provide high re performance also to non-re-rated, standard glass systems, as well as that FR glass curtains with water lms can o er high re performance, but limited heat resistance, or that the AFP systems can anticipate thermal shock failure in glazing windows and enclosures, leading them to premature collapse.

Existing Experimental Research on Glass
Properties under High Temperatures e performance of glass under high temperatures under heating and re loading attracted the attention of several experimental research studies, since 1950s, due to the huge use of glazing panels in windows and fenestrations. Most of those investigations are related to thermal shock e ects in   Due to the intrinsic properties of SLS glass, it is in fact known that, as far as the service temperature increases up to T g , its response becomes progressively time dependent, with rapid increase of permanent deformations. Standard AN glass, based on [42][43][44], proved to o er a typical brittle-to-ductile (BTD) transition at high temperatures, with toughness enhancement compared to other glass types. e same BTD behaviour, however, was also observed to strongly depend on the imposed strain rate, with BTD and transition temperature increase with strain rate increase [48].  e elastic properties of standard glass at elevated temperatures have been extensively assessed by Rouxel [47], by accounting for experimental data available in the literature after 1950s, giving evidence of SLS glass' MOE sensitivity to temperature, as compared with other glass types (see Figure 4, with SLS oat glass labeled as "window glass"). Rather linear dependency and limited decrease can be observed for MOE values of SLS glass, as far as T does not exceed T g , while a subsequent abrupt loss of sti ness is shown.
Earlier experiments were also carried out both on SLS and BS glass components by Kerper and Scuderi [49], with careful attention for specimens including (i) chemically strengthened SLS glass, (ii) thermally fully tempered SLS glass, and (iii) thermally semitempered BS glass. rough the experimental study, glass laths with dimensions of 254 × 38.1 mm (6.35 mm in thickness) and 152.4 × 25.4 mm (2.54 mm in thickness) were considered. Given the (i)-(iii) specimen types and a reference temperature (0-560°C the tested range), almost stable MOE values were experimentally derived, even after sequential heating and cooling cycles. MOE values were generally found to be completely relaxed for temperatures higher than 400°C.
Close correlation can be observed with MOE variations in standard AN glass specimens, as derived from di erent literature sources (see Figure 5, where test results from Shen et al. [50] on monolithic SLS samples (75.43 × 14.80 mm the size, with 3.26 mm the nominal thickness) are also reported). e same Figure 5, nally, gives evidence of the typically increasing MOE values for BS specimens, as a function of increasing temperatures.
Worth of interest for structural design purposes is that Kerper and Scuderi [49] also assessed the resistance variations in SLS glass at high temperatures. In particular, no resistance losses were reported for temperatures up to 375°C (less than 5% losses, compared to room temperature), for thermally fully tempered SLS specimens. Substantial decrease of resistance was recorded only for temperatures higher than 500°C ( re exposure for several hours) and 550°C (15 minutes of re exposure). Chemically strengthened SLS glass showed indeed a pronounced resistance degradation with the temperature increase, up to 5% loss at 204°C (500 hours of re exposure), 5.8% at 260°C (500 hours), and 100% at 600°C (6 hours).
Following [49], a huge number of experimental studies related to SLS glass performance have been focused on thermal breakage assessment, being representative of the major cause of glass cracking for windows. e issue of glass thermal cracking and fallout has been rst raised in 1980s by Emmons [51] and other researchers [52,53], while in the last decades, an increasing number of experiments have been carried out on small-scale specimens, single glass panes, or double glass panes variably supported, under the e ect of re or heat radiation (see, e.g., [54][55][56][57][58][59][60]). Numerical investigations were, for example, proposed in [61-63],

Advances in Civil Engineering
giving evidence of edge and boundary condition e ects on the thermal response and breakage of standard window glass panes. Malou et al. [64] carried out thermal resistance experiments on 3 mm thick, SLS, AN glass specimens (15 × 50 mm their nominal size). A rather constant value was recorded for the tensile strength of glass, up to a temperature increase of 270°C (Figure 6(a)). Higher temperatures were indeed associated to a sharp decrease in the measured resistance (more than 50% the reference value at room temperature), giving evidence of thermal shock e ects and damage propagation in glass specimens, as well as of generally limited performances of AN glass. A rather smooth MOE decrease was also observed (Figure 6(a)).
Later on, Xie et al. [65] experimentally investigated the tensile resistance of SLS, AN glass specimens at high temperatures. Quasi-static tensile tests were carried out on small specimens, with thickness comprised between 4 mm and 12 mm (2 mm the di erence between each set of specimens). Test repetitions on specimens with the same geometrical properties were carried out at 25°C and 200°C, where the critical breakage resistance was derived as the rst cracking occurrence. In Figure 6(b), evidence of such test results (average values, with minimum and maximum values for each series) is provided. In accordance with [65], a negligible decrease of resistance was noticed for specimens exposed to 200°C, compared to room temperature results, while higher sensitivity was observed especially to glass thickness (Figure 6(b)).
Worth of notice that, as far as di erent literature references are examined (see, e.g., [66]), even counterposed experimental ndings can be derived, giving evidence of a typically high scatter and sensitivity of glass thermal resistance to elevated temperatures, hence suggesting further testing and investigations at the material level.
Moving from the material to the assembly level, such a need of further experimental assessment and investigation can be further perceived.
Experiments related to the thermal breakage of speci c glazing systems under re loading have been in fact carried out only recently, that is, for double glazing units [58] or pointxed glazing panes belonging to curtain walls [60]. In the case of point-xed panes, for example, a high sensitivity of thermal breakage (i.e., time of failure and crack pattern) was typically observed, based on the position of point connectors (see an example in Figure 7). e actual performance of such a kind of specimens-as expected from the examined boundary conguration-proved to be strictly related to combined thermal exposure e ects as well as to mechanical loading (i.e., self-weight of point-xed panels, leading to additional stress peaks close to the holes), hence requiring a detailed investigation of both combined aspects.
In this regard, Chen et al. [17] studied the thermal breakage performance of standard AN windows, under the e ects of combined thermal loads and wind pressures. Steel frame-supported, monolithic 0.6 × 0.6 m panels (6 mm their thickness) were subjected to a reference re loading and various levels of wind pressures (up to 11 m/s the wind velocity on glass surface). Glass cracking, in some case, occurred together with fallout of samples from the supporting frame. Test results (15 specimens in total), however, generally proved that the rst cracking time markedly decreases with increasing the imposed wind pressure; that is, mechanical loads (wind, in this speci c case) can highly accelerate the failure of thermally loaded glass systems. As a result, detailed investigations inclusive of combined thermal and mechanical loads well representative of the actual loading and boundary con gurations for the examined structural glazing system should be generally carried out.
In the case of LG systems, for example, the thermal performance of interlayers of common use should be properly taken into account. In this regard, Debuyser et al. [16] investigated the behaviour of monolithic and triple-layer LG specimens composed of standard AN glass, under the e ects of radiant heating. Nominal thicknesses of glass panes of 6 mm, 10 mm, and 15 mm were taken into account, being bonded together in LG sections by PVB or SG layers (0.76 mm or 1.52 mm the thickness of interlayer foils). Low-E coated, monolithic specimens were also included in the set of experiments. Both radiant and transmittance tests were carried out, giving evidence-in accordance with earlier research e orts-of the relatively limited resistance and low thermal performance of AN glass specimens, due to the premature occurrence of thermal cracks as well as due to the poor thermal reaction of bonding interlayers (in the case of LG specimens). Critical design issues were also emphasized, by taking advantage of a 1D model able to capture the actual thermal response of the tested specimens. ermal properties of PVB and SG foils, up to 340°C, were also reported ( Figure 8). Test results collected in [16]-even limited to maximum temperatures of 340°C-generally showed a close correlation with past literature references for SLS glass [67,68] (Figure 8). Worth of interest is also the thermal characterization of PVB and SG foils.

Existing Experimental Research on Glass Systems and Assemblies
Although the relatively large number of experimental studies focused on the thermal performance of glass as Rouxel [47] Kerper and Scuderi [49] Shen et al. [50] Figure 5: MOE variation as a function of temperature, for SLS annealed and BS glass types. 8 Advances in Civil Engineering a constructional material, limited literature e orts are still available on the re performance of full glass systems and assemblies (see a selection in Table 2).

Glass Walls, Facades, Enclosures, and Windows.
Glazing enclosures and walls attracted the attention of researchers especially during the last years, to assess the re performance  [64], and (b) dependency of thermal shock resistance to glass thickness (in gray italic, the number of tests for each thickness), in accordance with [65]. of novel FR solutions in place of standard glass. In doing so, the actual boundary and loading con gurations were properly taken into account in de ning the test setup and methods, so as to reproduce the testing conditions of fullscale specimens as a part of full buildings and complex systems. Glass enclosures designed for an extension of Washington Dulles International Airport (automated train system for passengers) were tested under re conditions in 2007, as reported by Mejicovsky [69] (Figure 9(a)). Glass panels with 3.8 × 3.6 m high dimensions and supported by interior steel frames (4.7 m the average bay width) were investigated. Special details were de ned (even using standard, non-rerated materials), so as to o er appropriate redundancy to the glazing system, even in the case of an accidental event. To this aim, a special mock-up was also designed, so as to simulate the actual re-loading condition for the glazing enclosure.
Glass elements for overheads and walls were designed in the form of LG sections, composed of two 10 mm thick FT glass panes bonded by a 1.52 mm thick PVB interlayer. Glazing joints were then realized by means of silicone rubber setting blocks and structural silicone sealant joints (Dow Corning 995 ™ type), while the glazing channel and edge trims consisted of minimum 3 mm thick stainless steel. e re test was stopped after 35 minutes of exposure, with temperatures in glass over 250°C (up to 400°C in the last 5 tests of the experiment). Postexamination of the glazing system revealed no cracking or dislodgement of components, but localized melting and o -gassing of the PVB interlayer was observed, in the form of small bubbles and/or partial delamination (see details of Figure 9(a)).
A full-scale standard re test according to EN 1363-1 and EN 1364-1 was reported by Machalická et al. [70] for a glass wall supported by an aluminum frame. e wall (3.475 × 4.57 m the overall size) consisted of three special FR, LG panels, gel-lled type (1.4 × 4.5 m the size of the central LG panel, 1 × 4.5 m for the lateral panels). Small gaps between adjacent glass panes were lled by means of special FR tapes and sealants. Collapse of the glass wall occurred after 49 minutes of re exposure, with maximum temperatures in the order of 150°C (Figure 9(b)).
Both double LG panels composed of special tempered glass, gel-lled type (SAFTI SuperLite II-XL ™ type, with 19.05 mm the total thickness), and monolithic glass panels (6.35 mm the thickness) were assembled together to obtain the glazing wall tested in [71]. Such a glass wall (two glass panels for each cross-section type) was frame-supported and had overall dimensions of 2.42 × 2.42 m. Fire experiments gave evidence of limited re performances of simple glass panels, compared to FR components. Monolithic glass panes in fact were characterized by premature fallout from the supporting frame (i.e., 3 minutes after ignition, with 0.8 seconds elapsing between the occurrence of rst cracks in glass and the nal fallout of monolithic panes (Figure 10(a))). Double LG panels, in contrary, remained intact up to test conclusion (>30 minutes), without visible cracks or failure mechanisms close to connections with the framing system. is is not the case of experimental studies carried out by Yang et al. [40] that tested monolithic, FR glass panels under standard re curves. e experimental study gave in fact evidence of major issues deriving from framing systems and related connections. Even the FR glass panels  proved to o er high performances under re loading; in particular, loss of integrity was observed to have origin in the glass-to-metal frame connection detailing ( Figures  10(b) and 10(c)). Further re experiments and numerical investigations on glass facade systems can also be found in [78][79][80][81][82], with evidence of the performance of speci c glass system typologies, including cable-net systems and inclined facades.

Retro tting and Enhancement of Standard Glass
Windows. Several types of protective lms able to improve the re performance of existing and novel standard windows are available on the market. Although these coatings do not a ect the room temperature elastic sti ness and resistance of a given glass pane to retro t, the same lms can be bene cial in delaying high-temperature e ects, hence resulting in increased FR performance.  [74] LG, frame supported F/n.a. n.a. n.a.
Davis (2013) [75] LG, frame supported  Figure 9: Glass enclosures under re. (a) Evidence of partial delamination, as reported in [60] and (b) full-scale glazing wall tested in [70]. In evidence, it is possible to notice the loss of wall integrity and initiation of gel melting, after 49 minutes of re exposure.
Koudijs and Csoke [72] rst gave experimental evidence of the high potential of protective lms for traditional glass systems, with careful consideration for the improved re performance of windows composed of SLS glass, by taking into account a case study building in Rotterdam (NL) (Figure 11(a)). Low-E coatings were interposed within traditional double glazed units, by assessing the e ects of their position (i.e., interior cavity face, etc.) under re loading. e integrity of window samples was ensured for 27 minutes in the case of AN glass but increased up to 60 minutes in presence of HS glass panels, hence giving evidence of the potential re performances of traditional insulating systems inclusive of special coatings.
Misawa et al. [73] also tested the e cacy of special refractory lms, basically intended for application on the interior side of existing standard windows. e typical specimen consisted in a double glazed unit, Low-E coated, with 1 ×1 m the reference size. ere, a novel refractory lm was attached to the interior face of the glazed unit (i.e., on the glass surface expected to be exposed to re). e refractory lm consisted of (i) an external polyEthylene terephthalate (PET) lm bonded to (ii) an ultraviolet protection layer and (iii) a silicate soda-based material layer (1 mm its thickness) (Figure 11(b)). 12 tests were carried out in total, including variations in AN glass supplier, thickness (8 mm or 12 mm), and specimen size (30 × 30 cm, 100 × 100 cm, and 94 × 94 cm) as well as Low-E lm surface of application and comparisons with clear  Figure 11: Experiments on coated glazing windows, as reported in (a) [72] and (b) [73], schematic cross-sectional view. uncoated specimens. All the experiments proved the high e ciency of refractory lms, allowing Low-E double glazing units to achieve minimum 20 or 30 minutes of re performance, as required for FR windows.
Although the mentioned research studies generally proved the potential and e ciency of such special coatings to enhance the FR of glazing systems, several aspects should be still assessed. Wu et al. [83], for example, experimentally investigated the high-temperature performance and thermal degradation of protective layers of common use for glass applications, giving evidence of gas emissions when exposed to re. [74] reported on re tests carried out on glazing oors belonging to the subway station "Olympiapark Nord" in Munich (Germany). In the ceiling tunnel, overhead glazing accessible to person steps was made accessible by means of several openings (5 × 3.5 m the size). FR requirements were taken into account, as a possible con guration deriving from accidents (i.e., trains burning in the tunnel). Special multilayered safety LG panels were designed and tested under re, in order to ensure appropriate safety levels ( Figure 12). No test results and re-performance-related data are available in the literature, however.

Glass Floors and Overheads. Siebert and Maniatis
Similar re experiments are summarized also in [75], referring to the glazing oor panels designed, in 2011, to be installed at a height of 130 m in the historic Blackpool Tower (UK), as a part of an ongoing refurbishment project. A full-scale re test was carried out (Figure 13), with 4.42 × 3.8 m the overall size of the ooring system. e mild steel-treated frame and related gaskets were also included within the test setup, so as to assess the re performance of the full glazing system under its actual restraint con guration.
Aiming to ensure appropriate structural integrity as well as overall performances in the case of re accidents, the typical glass panel (3 × 1.1 m the maximum size) consisted of a LG section, three 10 mm thick, FT glass layers, bonded together to a 23 mm Pilkington Pyrostop ™ , by using a special liquid composite bonding (Koediguard ™ type). Steel-framing elements were also preliminary treated with intumescent coatings, mineral wool, and re check boards. Additional expansion joints were nally included in the setup, in order to prevent buckling in the supporting frame members. e re experiment was carried out in accordance with EN regulations, assigning to the ooring system a standard re curve and a simultaneous mechanical load, being representative of crowd (5 kN/m 2 , distributed uniform pressure or 4.5 kN point load (50 × 50 mm foot print), resp.). e re performance exceeded 60 minutes of integrity (test stopped after 68 minutes), with maximum recorded temperatures of 67.6°C and no evidence of debonding or failure. Compared to re insulation requirements (EN 1363-1), where temperature rise should not exceed 140°C (Section 2), a tolerance of > 120°C was hence ensured (12°C the ambient temperature during the test). Posttest de ection of 16.5 mm was measured (with 1/175 the roof span de ection limit value provided by standards).

Glass Beams.
Limited experimental literature background can be found for structural glass elements under re loading, due to the relatively recent demand of vulnerability assessment and protection for these systems. Veer et al. reported in [22] on a set of bending test comparative results, experimentally obtained from monolithic and LG beams under re, to assess the e ects of intumescent coatings. e re loading was imposed in the form of a constant ame at 650°C, with a xed distance from the beams' lateral surface ( Figure 14). Various beam geometries (40 mm × 400 mm the overall size) were tested, including specimens with di erent thickness and standard SLS glass types, such as (a) 6 mm thick, AN glass; (b) 6 mm thick, chemically toughened glass (120 MPa the initial stress); (c) 3 mm thick, chemically toughened glass, laminated with 1 mm polycarbonate (PC) foil (3 glass layers + 2 PC lms); (d) 3 mm thick, segmented, chemically toughened glass, laminated with 1 mm PC foil (3 glass layers + 2 PC lms). Compared to C, glass layers were bonded to PC foils in an overlapping pattern; (e) An LG beam (as in the case of C and D type specimens), including insulating cavities on the external sides.
All the (a)-(e) con gurations were tested both without and after the application of intumescent coating (FlameGuard HCA-TR ™ paint type).
A conventional four-point bending test setup was used for these beams. Experiments with no external mechanical loading as well as with additional weights (with up to 24 MPa the corresponding midspan bending stress) were carried out on monolithic AN specimens (type A) ( Table 3). e overall experimental investigation gave evidence of some important aspects, as, for example, the potential safety level of structural glass beams under re.
However, critical aspects were also emphasized for the same specimens, like, for example, in the case of segmented beams (type D, see Figure 14(c)). Some preliminary FE simulations were also reported in [22], giving evidence of the temperature distribution and related stress e ects for the examined beams. Worth of interest is in fact that such preliminary FE models gave evidence of temperature peaks in the adhesive layers providing structural bonding between the glass segments, hence emphasizing the crucial role of detailing.
Bokel et al. [76] later explored similar glass beam specimens, by taking into account the same overall geometrical features and test setup presented in [22]. e novel aspect was represented by testing LG beams composed of special FR glass (i.e., Pyroguard ™ type as well as LG beams composed of 3 SLS glass layers, with epoxy lms acting as interlayers for all the specimens). As a general outcome of the experimental investigation, epoxy layers were found to start charring after few seconds only, with limited re performance of the beam specimens. An almost comparable behaviour was observed for all the beams, both composed of special Pyroguard layers or not, hence giving evidence (besides the limited number of tests) of the need for further extended investigations.
Louter and Nussbaumer [77] performed full-scale experimental tests on LG beams composed of standard glass layers. Di ering from [18], a standard re curve was considered for loading onto the oven, in accordance with EN regulations (Section 2). rough the experimental study, 3 full-scale beams were investigated. Given the same overall dimension of beams (1 m × 0.1 m), variations were accounted in terms of glass type (AN, HS, and FT, resp.). e reference cross section consisted of 3 SLS layers, 10 mm in thickness, bonded together by SG foils (1.52 mm in thickness).
A four-point bending test setup was considered, with end supports protected from re exposure, and the re loading was assigned together with a simultaneous, constant mechanical load taking the form of 115 kg at the midspan section. Given the limited stress e ect due to the assigned mechanical load (with maximum tensile stresses in the order of 5 MPa at beams' midspan), the specimens proved to o er a rather stable behaviour under re, for >40, >45, and >50 minutes in the case of AN, HS, and FT beams, respectively, up to collapse ( Figure 15). As a general observation from such a kind of tests, the interlayer foils started melting and leaking o from their position after few minutes of re exposure only; hence, the SLS glass panes behaved as almost fully uncoupled layers. On the other hand, protecting the beam ends from re allowed to avoid premature collapse mechanisms.

Summary and Conclusions
In this paper, a state of art on structural glass systems under re loading was presented, with careful consideration for current design methods and issues as well as experimental research e orts. Besides the continuously increasing use of glass in buildings as a constructional material able to interact with and/or replace materials of traditional use, the actual behaviour of structural glass assemblies, in general, currently requires further investigations, as well as the application of speci c fail-safe design rules. is is the case of glazing systems under ordinary loads, but especially of extreme loading conditions, as, for example, re accidents.
As shown, the intrinsic features of glass and its interaction with other components (i.e., framing systems, boundary details, etc.) make glazing systems highly vulnerable to temperature variations, as well as combined effects of thermal and mechanical loads, hence requiring multidisciplinary approaches in their design. In doing so, appropriate structural safety levels should be in fact ensured in combination with multiple aspects, such as transparency, aesthetics, and lightening requirements.
At the material level, in particular, a wide set of experimental research can be found in the literature, aiming to assess major e ects of high temperatures on MOE, tensile resistance, and thermal cracking of standard glass. Most of these experimental outcomes are in rather close agreement as far as the MOE variation with temperature is considered. But when di erent literature sources are accounted, however, test results can also give evidence of high scatter in the observed trends, as, for example, in the case of glass thermal resistance (Section 4). In addition, while most of glass applications in building consist of laminated or insulated glass systems, few experimental studies only are actually available to characterize the thermal e ects in interlayer foils at high temperature.
As far as the attention moves from the material to the system and assembly levels, a relatively wide set of experimental investigations can also be found in the literature, with careful consideration for the re performance of various Bene cial e ect of intumescent paint, even with signi cant thermal damage; central 7 cm PC melted (visible after 10 min) * Mechanical load included; I integer beam; BR broken; C cohesive failure between glass segments and PC foils; T max maximum temperature monitored on the beam backside, at the end of the experiment; n.a. not available.
Fire-protected beam supports Figure 15: LG glass beams tested by Louter and Nussbaumer [77]. Typical failure con guration. typologies of glass systems inclusive of a multitude of boundary con gurations, re exposure patterns, and glass types (standard and/or FR glass). As a common aspect of such experimental investigations (Section 5), connection details and restraints generally proved to have a key role in the overall observed responses, both for frame-supported and point-supported systems. Generally speaking, glass enclosures, walls, and beams proved-in most of the cases-to o er rather stable performances under re loading, even composed of standard glass only, but requiring further extensive testing and assessment with special care for supporting details.
Finally, literature e orts have been spent in the last years also to assess the potential and e cacy of special coatings and lms for the retro tting and protection of existing glass windows and systems in general. In accordance with earlier observations, such solutions generally gave evidence of major bene ts for uncoated glass specimens, but careful consideration should be still spent to properly optimize their potential.

Conflicts of Interest
e author declares that there are no con icts of interest.