Pre- and Post-Failure Experimental Bending Analysis of Glass Elements Coated by Aged Anti-Shatter Safety Films

The main goal of Anti-Shatter Films (ASFs) applications for structural glass is to create a barrier able to keep together fragments and minimize risk after any impulsive or static load that could lead glass to cracking. The influence of ASF properties on the flexural strength of coated glass elements is thus a relevant topic for safe design purposes, but still little investigated. To this aim, an experimental material investigation is presented in this paper, in order to achieve a good knowledge of common ASFs from a chemical point of view. Moreover, the deterioration of mechanical and adhesion characteristics for ASF samples subjected to different environmental conditions and accelerated ageing is also investigated, so as to simulate the effects of long-term exposure to high humidity (HU) or high temperature (HT). An experimental campaign carried out on 20 small scale ASF-coated glass specimens is finally presented, based on a three-point bending (3PB) test setup. The out-of-plane bending response of unaged or aged samples is performed by taking into account two different displacement-rate levels, to assess their performance and bending capacity under steady-static or impulsive loads. In both cases, the attention is given to the characterization of elastic and post-failure performances. Finally, support for the interpretation of experimental outcomes is derived from a simplified theoretical model of composite beam with partial connection, in order to estimate the shear stiffness of ASF adhesive components in the elastic stage.


Introduction
Anti-Shatter Safety Films (ASFs) are typically used in practice to mitigate the consequences of glass breakage. The first applications can be found in the early 1970s by British Government, in order to lessen the effects of terrorist bombing. Thanks to the impressive results, the use of ASFs began to expand in different fields of application. Meanwhile, the improvement in choice of materials and / or manufacturing technologies were developed by companies interested in the versatility of the product. Nowadays, there are several anti-shatter glass films on the market, almost all transparent offering invisible protection in case of an accident, thus against shattering, breaking, explosions or vandalism. Therefore, it is crucial to assess and quantify the efficiency and benefits for different application surfaces. Although this kind of commercial products have passed several standard certification tests as reported in the Guidance Note (2013), the poor literature in experimental experiences that provide a measure of the increasing glass system strength (Ahn 2019;Kojima et al. 2021;Memari et al. 2004;Van Dam et al. 2014) should be improved. Generally, to reduce or avoid risks structural robustness can be provided in glass constructions by using redundancy at different levels. Several studies were aimed at mechanical characterization of laminated glass plate subjected to bending loading with different interlayers (Serafinavicius et al. 2013) or environmental temperatures (Antolinc 2020) or progressive damage (Biolzi et al. 2010). The main advantage of laminated glass is its capacity to act as safety glass, due to section redundancy, in different areas such as for windshields in the automobile industry (Batzer et al. 2007) or classical façade components (Weggel et al. 2008). Therefore, most of the studies conducted on the bending behaviour and the impact response of glass have focused on laminated sections, due to their ability to remain substantially intact (Castori et al. 2017;Iwasaki et al. 2007;Sable et al. 2017;Timmel et al. 2007). Alternatively, the major disadvantage of architectural float glass is the lack of residual strength, so generally fracture corresponds to sudden collapse.
In this regard, the present experimental study is focused on the investigation of the out-of-plane bending behaviour of aged monolithic glass specimens fitted with ASFs, analysing both the elastic and the post-fracture phases. Two accelerated ageing protocols are employed, exploring the influence of high temperature and humidity on the adhesion performance, as also in accordance with (Butchart & Overend 2017). The resistance improvement for glass element is a fundamental topic in a vulnerability analysis with regard to natural causes, such as earthquakes, hurricanes, etc., and even accidental events. The assessment of ASF-coated glass behaviour in out-of-plane bending is hence carried out in this paper via three-point bending (3PB) test setup, with both a quasi-static or impact regime. For the plane elastic problem, theoretical expressions are developed by considering shear phenomena at the interface between glass substrate and PET-film. Finally, the basic material characterization and mechanical capacity for the tape and ASF-coated samples is discussed.

Theoretical analysis
Generally, the mechanical coupling of two or more structural elements in out-of-plane bending takes place through the use of special connection systems (i.e., composite slabs, columns, beams, etc.). These connections, both discrete or continuous, have the task of resuming the sliding effort between the linked elements, in order to transmit an axial effort between them. In the field of structural glass applications, typical examples include adhesive linear bonds for hybrid beams (Louter 2011;Firmo et al. 2020;etc.).
In the same way, the contribution of ASF in the out-of-plane bending behaviour of coated glass elements can be taken into account thanks to the presence of PSA, whose possibility of providing a residual resistance is analysed in this study. The reference composite system can be seen in Fig. 1, where subscript "1" and "2" are related to the glass element and the PET-film respectively.
In doing so, a basic important approximation is made in considering the absence of peeling phenomena until the complete detachment, which corresponds to the loss of strength in the test procedure. Since the efficiency of the composite structural element is higher when the shear stiffness of the connection is high, the smaller are the displacements relative to the interface of the two elements. Typically, the real connection behaviour is intermediate between zero and infinite stiffness ( Fig. 1). In the former case, the two elements are simply in parallel; considering Bernoulli's hypothesis valid, the pane sections of the individual component beams are preserved. Instead, in the latter case, the two bonded components are rigidly coupled together and the overall section of composite system deforms as homogeneous one, with no sliding at the interface. Additionally, the constitutive law of these connectors is rarely linear, and may suffer for degradation due to ageing.
According to original Newmark's theory (1951), for intermediate conditions of partially rigid connection and relative sliding as in Fig. 1, the real behaviour of the composite beam can be reduced to the response of two elements bending in parallel and connected by a deformable connection with elastic-linear stiffness K. The solution by Newmark is valid for a simply supported beam made up of a maximum of two different layers. Furthermore, in this discussion several hypotheses are considered, such as (i) uniformly distributed connection along the span, (ii) constant geometries / section properties, (ii) linear elastic behaviour of materials and (iv) same curvature for elements "1" and "2". Adaptations of Newmark's analytical model for laminated glass elements can be found for example in (Amadio & Bedon 2010;Amadio & Bedon 2011) for extension to buckling-related aspects.
On the basis of above assumptions, and through some simplification and integration steps, it is possible to arrive at the equation of composite beam as in Fig. 1 (1) with: M is the bending moment to which the beam is subjected, t ad is the stiffness of the adhesive and G ad is the tangential stiffness of PSA. The coefficient α is function of the equivalent axial stiffness EA*, the bending stiffness EJ full (determined in the hypothesis of rigid shear connection between elements "1" and "2") and the bending stiffness EJ abs (as in absence -or very weak -shear connection). These equivalent terms can be calculated as: * = 1 1 × 2 2 1 1 + 2 2 (4)

Experimental program
The present study aims to analyse the out-of-plane bending response of ASF-coated glass elements by investigating the influence of temperature and humidity on its key mechanical parameters. The experimental campaign consisted of 20 three-point bending (3PB) tests on glass elements fitted with ASF and 15 uniaxial tensile tests on PET.
In particular, the flexural bending tests covered two imposed deformation rates, thus resulting in low speed (0.5 mm/min) and high speed (1 m/min) experiments respectively, so as to analyse the response under impulse. Moreover, to characterise the influence of individual components, a study was carried out on the mechanical characteristics of PET to quantify the effects of temperature and humidity variations on the elastic modulus of this material.

Materials characterization
The ASF material under investigation was selected to coincide with a commercially available multilayer film. Special attention was paid to detect the differences in tensile behaviour after ageing (film only) and in out-of-plane bending behaviour (when coupled to glass). All glass elements were provided by the same factory and manufactured by annealing process.
The examined multi-layer tape is sketched in Fig. 2. As shown, the typical ASF is composed of three different layers (0.11 mm in thickness) bonded together by an interposed adhesive, and a PSA adhesive to ensure contact with glass substrate. The latter is usually protected by environmental conditions by means of a removable release liner. The layer material as well as the adhesives were investigated by Fourier Transform Infrared spectroscopy (FT-IR) and Differential Scanning Calorimetry (DSC). FT-IR is a technique that allows investigating the chemical nature of materials by analysing the vibrational bands associated to different functional groups. DSC is a thermal analysis technique that evaluates the heat exchange difference between sample and reference as a function of temperature and time. In the case of crystalline or semi-crystalline materials, information on the melting and crystallization temperature and enthalpy associated with phase change can be obtained.
As shown in Fig. 3, FT-IR spectra of layer-1, layer-2 and protective liner (acquired on both sides) and internal side of layer-3 show vibrational bands consistent with polyethylene terephthalate (PET) (Socrates 2004). FT-IR spectrum of external side of layer-3 show different vibrational bands, overlapping those typical of PET: this are consistent with an amine. Surface treatments of PET with such compounds are often used as stabilizers or UV absorbers (Aflori et al. 2007;Rusu et al. 2008). Fig. 3: FT-IR spectra of layer 1,2 and 3. Layers 1 and 2 (both sides) and layer-3 (internal side) spectra show good match with PET vibrational bands (vertical lines). Spectrum of layer-3, external side (blue curve) shows some minor difference respect to PET vibrational bands due to extra vibrational band ascribable to an amine surface treatment. FT-IR spectra of the adhesive between the layers and that of PSA adhesive are identical (i.e., Fig. 4), pointing out that the adhesives share the same chemical composition. Vibrational bands identified CH 3 , CH 2 , C-O-C and C=O groups; the overall spectrum is consistent with that of a poly ethyl acrylate or poly allyl acrylate (often used as PSA adhesive), see for example (Mapari et al. 2021;Marquez et al. 2020).
FT-IR spectra of layers and adhesive after ageing processes do not show any difference respect to unaged sample spectra. This implies that no chemical degradation (oxidation) took place on PET layers and that no degradation or cross-link reaction occurred in the adhesive during the ageing processes.  (Liu et al. 2013;Piccinini et al. 2013;Son et al. 2003) and their melting temperature is consistent with this secondary melting peak.

Test specimens and setup
As shown in Fig. 6, the reference sample size of 100 mm х 40 mm x 6 mm and 100 mm х 40 mm x 0.35 mm (length x width x height) were considered for glass and ASF respectively in 3PB tests. Details of the setup of the bending tests are also shown in Fig. 7. Tests were performed with a Shimadzu AGS-X universal testing machine; the span between the lower supports was set to 64 mm.  A crosshead speed of 0.5 mm/min was used for low speed tests (10 specimens), 1 m/min for high speed tests (10 specimens). This latter high displacement rate was applied in order to evaluated the flexural strength before cracking and its residual value thanks to the presence of ASF. However, in both cases the crack initiation and propagation process in the glass was very fast and rather difficult to monitor. The span between the supports was 64 mm. The diameter of 5 mm characterized both loading apparatus, equipped with a load cell, and the steel supports along their entire width.
The experimental data (force, displacement) were automatically recorded for each time step using a data acquisition system, where the load was applied at the midspan through a fixed displacement rate but different for two sets of specimens: • low-speed rate for HU-i and HT-j, with i=1 to 5 and j= 11 to 15; • high-speed rate of for HU-i and HT-j, with i=6 to 10 and j= 16 to 20.
Regarding tensile tests, specimens of clean PET-film with the same geometry (200 mm x 25 mm x 0.34 mm) such as represented in Fig. 8, were tested by means of tensile tests with a Shimadzu AGS-X universal testing machine, with a crosshead speed of 25.4 mm/min at room temperature (Fig. 9). These were tested to characterize the mechanical properties of the tape material without any interferences of the PSA; they were obtained after removing the PSA with a suitable solvent (acetone, specifically chosen to be able to dissolve PSA but be inert towards PET). Moreover, an interesting investigation on tensile properties on each layer was performed after separating the layers composing the PET tape and testing them individually with the same parameters (crosshead speed of 25.4 mm/min).

Ageing procedures
Two accelerated ageing procedures were investigated in order to simulate the effects of long-term exposure to high humidity (HU) and high temperature (HT) environments, according to EN 12543-4 that is specifically developed for identification of glass laminates (with PVB, EVA or SG) durability. In HU procedure specimens were held in a vacuum box in a vertical position over water for a period of 2 weeks within the limit of 50°C, while in HT procedure specimens were heated in a convection oven to 100°C for a period of 120 minutes. Subsequently, the tests were performed within a limit of 3 hours.

Experimental results and discussion
Stress-strain curves of tensile tests performed on clean PET material are shown in Fig. 10. Stress-strain curves of each layer are shown in Fig. 11. The evaluation of the tensile test of the PET tape was preparatory to the analysis of the flexural response of the composite section (glass + adhesive + PET).
The main goal of this test was the comparison of results in terms of mean modulus of elasticity (E 2 ) between the three sets, that is unaged and aged specimens (based on HU and HT ageing procedures respectively). The load-displacement curves of tensile tests for PET tape showed a very similar qualitative behaviour before and after aging; as shown in Fig. 10 (b), there was no discernible difference in E values (3.3 GPa the average), giving evidence of high consistency with previous analyses on similar components (Mattei et al. 2022).
With respect to the individual films, it can be noticed in Fig. 11 that the average moduli of elasticity of the single layers are slightly greater than the actual value of the overall unaged section. The load-displacement curves of the low-speed 3PB tests, carried out on "glass+film" system, are depicted in Fig. 12. The first part of the curves was elastic-linear, generally until the maximum force value representing the glass failure (F break ) was reached; after this point, an abrupt reduction in strength was detected, followed by a segment in which a residual value of strength (F residual ) was recorded, most likely due to the presence of the ASF. Finally, the last portion of the load-displacement curves, after the second sudden strength loss, corresponds to the film's contribution alone following the complete detachment of the adhesive layer on one or both bottom sides of specimens caused by the opening of cracks. Monolithic brittle materials, such as annealed glass, are known for the typical sudden failure, which usually occurs in the elastic region. The benefit of having a residual strength, provided by safety films, is one of the main aspects which lead to the choice of using this type of tool to increase the postbreakage behaviour of float glasses. In order to evaluate the influence of ASF on the bending behaviour of the glass element, an analysis of F break and F residual values was performed, and the relative results Multilayer Film Emean were summarized in Table 1. The variation observed between the maximum load values before the fracture, represented by the standard deviation reported, is due to the non-uniform behaviour of glass that depends by many factors, such as manufacturing technique, edge defects, chemical composition and crack pattern (i.e., Fig. 13). As a consequence, it should be noted that the crack distribution typically concentrated around the midspan but the position of initiation cracking point and the number of cracks change. However, HU-1 curve was neglected in the evaluation of the parameters reported in Table 1 because it involves singular values, also supported by a different crack pattern (i.e., Fig. 12(a)). The findings summarized in Table 1 are reflected in the Fig. 12: the exposure to HU ageing led to an increase of interfacial adhesion and, consequently, of ratio F residual /F break with a corresponding increase in displacement at failure. However, the exposure to HT ageing provides better results than the previous one in terms of residual strength and maximum displacements until the overall collapse.  Based on the theoretical background by Section 2, a numerical parametric analysis was performed in order to calculate the tangential stiffness of the adhesive layer related to the elastic phase, by imposing v(L 0 /2)= v exp . Where v exp represents the displacement recorded during the 3PB tests by testing machine and v(L 0 /2) corresponds to the theoretical solution by Eq. (1) in the same region.
Moreover, Fig. 14 shows tensile stresses evaluated at the bottom surface of the glass, σ 1,max (in the hypothesis of rigid connection) and at the bottom surface of the film, σ 2,max , in the midspan section corresponding to the maximum bending moment FL span /4: with n the ratio between the modulus of elasticity of PET and glass (n= E 2 /E 1 ). The investigation of the tests results was done by discussing the effects of the ASF on the mechanical behaviour of monolithic glass. A further analysis on the post-breaking branch was focused to build an equivalent numerical model able to calculate the residual strength and deflection taking into account that the section does not crack but that the tangential stiffness of the adhesive decreases following a power-law, as reported in Fig. 15. Although the trend of the two curves appears to be the same, the For the high-speed 3PB tests, the load-displacement curve presents a completely different shape as shown in Fig. 16. In this case the analysis of the results was performed in terms of equivalent strain energy, or absorbed energy in impact bending, computed as the area under the load-deflection curve up to the peak force, related to the onset of the first crack. This parameter corresponds to the toughness (K el ) of the linear region reported in the last column of the Table 1 and in Fig. 17 (a). Furthermore, another benchmark for comparison of the bending behaviour in different ageing conditions can be F max as reported in Fig. 17 (b). It can be noted that the HT ageing procedure provided a greater load-carrying capacity in the elastic and post-brakeage phases for low-speed rate 3PB test, and greater tensile toughness and maximum force in case of v= 1 m/min.

Summary and conclusions
As noticed, the maximum bending force and midspan deflection at failure under three-point bending test highlights the uncertainties involved in the glass breakage process. Both accelerated ageing procedures considerably affected flexural strength of coated glass specimens. Therefore, the behaviour of specimens after cracking is totally influenced by the PET-film adhesion characteristics, thanks to which the remaining fragments held together by the tape express a little residual strength and a deformation at collapse, recorded at L 0 /2, greater than twice that related to cracking. In particular, comparison of the environment influence showed that the selected conditioning methods affected the adhesion performance and thus the bending response of the "glass+ASF" composite system.
By a chemical point of view, the pre-conditioning with the accelerated ageing procedures does not involve any modification in the components. Moreover, there are no major differences in failure mechanisms between the four sets of specimens. In qualitative terms, the force-displacement response was affected by viscosity property of the PSA, and in particular of the tackifier resin which improve its wettability to the adherent and accelerate the adhesive bonding to the substrate.
Of course, all ageing conditions investigated in this work caused a change in the post-breakage branch: the exposure to high humidity for two weeks provided about 0.1F break as residual strength, and the exposure to high temperature showed higher values of peak forces and just over 40% as a residue. In this latter case, the tape increased ductility feature beyond the linear-elastic phase. Conversely, with a high-speed rate tests flexural response didn't show a post-fracture curve. Therefore, the displacement-rate influence was assessed in terms of dissipated energy as strain and computed as the area under the curve up to the peak value. Furthermore, the ageing on PET-film mechanical parameter were found to be of negligible influence by mean tensile tests. This paper highlights the need for further study that take into account tangential and normal stress components at the interfacial surface by including the delamination process in the theoretical treatment. HU-6/HU-10 HT-16/ HT-20