A completely transparent, adhesively bonded soda-lime glass block masonry system

2.1The case study

The novel glass masonry system will be realized on the 10 × 12 m facade of the (Crystal House) (see Figure 3), designed by MVRDV and Gietermans & Van Dijk architectural offices. The completion of the project is expected to be accomplished in 2015. Located in the historic Pieter Cornelisz Hooftstraat in Amsterdam, the building had to follow strict planning regulations that required the new facade to maintain the same organization, rhythm and composition as the one of the previous 19th century building. To conform to these restrictions, yet to design an interesting facade, MVRDV came up with an ingenious solution. The Crystal House’s facade is an accurate reproduction of the previous building’s historic facade, yet with one great difference. Instead of brickwork, it is all made of glass: from the bricks to the doors and window frames, everything consists of clear glass. As the facade rises, normal clay bricks intermingle between the glass blocks, to create a smooth, gradient transition to normal brickwork on the top floor. The end result is a building that will stand out, and at the same time will naturally blend into the urban fabric of the historic street.

In principle, the regeneration of the brick facade using glass instead of clay blocks was structurally feasible due to the compression-loading scheme of a wall structure. Considering that annealed glass is usually ten times stronger in compression than tension, it is very suitable as a load-bearing material for such an application. As previously mentioned, the new facade of the Crystal House is composed of solid cast glass blocks that intermix with normal bricks towards the second floor of the building. More specific, the lower ten meters of the facade comprise mainly cast glass blocks. Only at the highest part of this area, the glass blocks intermix with conventional clay bricks in a limited zone until the first array of solely clay bricks appears (see Figure 3). Above this array, a steel beam covered with clay bricks is placed to support the upper, traditional brick facade. The beam is connected to the slab of the second floor, thus allowing for a thicker brick facade (with cavity) to elevate from that point up. At the current stage of research, epoxy is proposed to be used for the connection of the glass block facade with the conventional part of the construction, as well as for the bonding of glass blocks with clay bricks. Nevertheless, this might be modified, since the research for the final selection of adhesives between glass, bricks and steel is at the time of writing still continuing.

To reinforce the flat geometry of the massive glass wall against lateral forces and buckling that may occur due to induced wind eccentricity, four buttresses project inwards from the facade. In specific, the buttresses are 5.5 m tall and are formed by glass blocks interlocking to the ones of the facade, thus forming a continuous relief glass envelope (see Figure 5).

The glass facade, weighs approximately 25% more than a standard masonry facade of the same dimensions due to the higher density of glass compared to brick. This 25% difference of dead load necessitates a reinforced foundation.

Except for the use of glass bricks, the main difference between the old and new masonry system is that the glass wall’s thickness is covered by the width of one brick (210 ± 0.25 mm) instead of two, as in normal masonry (see Figure 5). This was specifically chosen to eliminate unnecessary joints that can affect both the structural and optical performance/clarity of the glass wall. Accordingly, to reproduce the isodomic modulus of the historic brick facade, all glass blocks have the same width (210 ± 0.25 mm) and height (65 ± 0.25 mm) but are cast in three different length sizes (105, 157.5 and 210 ± 0.25 mm).

2.2The glass masonry system

The architectural prerequisite was to obtain a completely transparent and at the same time structurally feasible solution. To meet this requirement, a combination of solid glass blocks and colourless adhesive is used for the fabrication of the glass masonry wall.

In particular, solid glass blocks were selected for their high compressive strength (typically over 200 MPa), which renders them able to carry the dead load of the wall without the need of any supporting structure.

To meet the aim of maximum transparency, a colourless adhesive needs to be used to bond the glass blocks together. The mechanical properties of the chosen adhesive are equally critical to the properties of the glass blocks, as in masonry structures only the properties of the whole structure are important. It is their combination and their interaction as one structural unit that defines the structural performance of the facade and not the individual strength of each element. More specifically the adhesive should:

An adhesive that meets all above demands is a one-component transparent, UV-modified acrylate, designed for exterior glass-to-glass bonding. This adhesive is photocatalytically cured and after hardening it becomes moisture- and water-resistant. The specific adhesive reaches its optimum strength when applied in a layer of approximately 0.1– 0.3 mm thickness. Figure 4 illustrates how a comparatively thinner or thicker layer critically affects the adhesive’s shear strength and thus the performance of the adhesive. This means that irregularities on the glass block surfaces can result in an uneven spread of the adhesive which will affect the load-bearing capacity of the wall by creating weak spots. Hence, to obtain the most favourable structural capacity, the glass blocks need to be fabricated with very high dimensional accuracy, to ensure an even and thin distribution of the adhesive. This accuracy is also required for architectural purposes. An inconsistent spread of the adhesive can result in visible gaps and bubbles. But most importantly, considering that the joints between adjacent blocks have virtually zero thickness, even a tolerance of 0.5 mm per block could create a sizeable offset in height or width of the construction. An essential difference between this adhesive system and a conventional mortar system is that mortar can allow for variations in brick dimensions, while this adhesive cannot.

As a result it was determined that the glass blocks’ size and flatness should be confined into a tolerance that can be covered within a homogeneous adhesive layer. This tolerance was found to be ± 0.25 mm in both flatness and dimensions. To verify this tolerance several architectural prototypes were made, on which structural experiments were conducted.

Glass blocks with such strict tolerances in size and flatness have never been asked for or produced before. In projects where metal substructures are used, sealant joints of considerably bigger thickness are used, which can take up the rougher tolerances. On the other hand, in the Atocha Memorial, the sole comparable structure, the overall cylindrical shell geometry that contributes greatly to the structure’s rigidity, allows for a tolerance range of ± 1 mm (Christoph & Knut, 2008, p. 114) without decreasing the structural capacity. However, in this case study the 10 m × 12 m dimensions of the facade and its flat geometry necessitate maximum masonry strength and consequently require optimum thickness of the adhesive.

Soda-lime glass was selected for the casting of the glass blocks, in contrast to borosilicate glass, which until now has been preferred for architectural applications. The reason is that borosilicate glass has a low coefficient of thermal expansion [3.2 – 4 *10⁻⁶/K]. As a result, it can resist thermal shocks more easily, reducing thermal expansion to half compared to soda-lime glass [9.1 – 9.5 *10⁻⁶/K] (Christoph & Knut, 2008, p. 113). But most importantly borosilicate glass has noticeably less shrinkage on cooling, leading to higher dimensional accuracy of the end product, eliminating most of the need for post-processing. However, in our case, the strict 0.25 mm tolerance requires post-processing even if borosilicate glass would be used. Yet, data from the standardized float and cast production of glass (see Table 1) show that soda-lime glass generally has higher compressive and tensile strength than borosilicate glass. The cast and float manufacturing processes of soda-lime glass lead to closely comparable mechanical properties. With proper processing, the chemical compositions are virtually identical, the glass element is controllably and slowly cooled, while the occurrence of defects such as air bubbles is limited. For example, cast glass objects from either soda-lime or borosilicate have virtually equal strength values compared to floated elements of the same type of glass (values derived from CES EduPack 2014). However, in cast objects of considerably larger thickness, the manufacturing process is more difficult to control and thus the number and size of defects increase. For this case, cast solid soda-lime glass blocks, the strength is expected to be less than given for soda-lime glass in Table 1, but still proportionally better than would be found for cast borosilicate glass.

The custom-made fabrication of the final blocks was assigned to Poesia Company in Italy. The massive blocks are manually cast in specially designed precision moulds coated with nickel, to produce components with smooth surfaces and remove the glass out of the mould more easily. Low-iron glass is used to produce completely colourless glass blocks. Initially, liquid glass, melted at approximately 1200°C is poured into the steel moulds (see Figure 6) and left to cool to 700°C. During this stage, the top face becomes significantly convex due to gravity. After the block is cooled down to 700°C, it is removed from the mould and is placed in an oven to slowly cool down from 700°C to room temperature. The cooling process is time and temperature controlled, lasting from 8 to 36 hours depending on the size of the block, in order to avoid thermal cracking and internal residual stresses. Indeed, measurements done by a Scalp 5 (GlasStress Ltd, 2014) device on specimens indicate that the residual stresses are less than 10 MPa – which is the residual stress limit that the Scalp 5 device can measure. The resultant block is placed in a CNC machine, which removes the convex top face and processes the block to the precise height required. Finally, the two horizontal faces of the blocks are polished to a smooth flat surface, to minimize any local projections of the glass surface that can lead to peak stresses if highly loaded.

Prior and parallel to structural testing, several architectural mock-ups of the masonry wall were built to study the visual performance of the system and to determine the minimum tolerances needed for architectural purposes. Initial research indicated that due to the low viscosity of the specific adhesive, the vertical joints of the blocks cannot be homogeneously bonded: the adhesive would flow downwards before it could be cured. Therefore, it was determined that only the horizontal joints of the blocks would be bonded.

Next, three successive wall mock-ups were made, using blocks of the final custom-made dimensions but with different tolerance ranges. The results are summarized in Table 2. The three mock-ups are shown in Figure 7.

From the results it can be determined that larger tolerances firstly lead to significant offsets in the height and width of the facade, secondly to open joints between adjacent blocks and thirdly and most importantly to an uneven spread of the adhesive that can greatly affect the structural performance of the wall. A satisfactory visual result is only achieved if tolerances of ± 0.25 mm are strictly met in size, rectangularity and flatness.

A fourth mock-up was then constructed (see Figure 8) with the desired tolerance on the blocks, where the construction method of the buttress can also be seen.

2.3Maintenance

The proposed system is practically maintenance free. Glass block is a durable material and generally is not subject to weathering (Dietrich et al., 1995, p. 166). In order to minimize the cleaning requirements of the facade, a spray of hydrophilic coating (e.g. Vindico (Vindico, 2014)) can be externally applied to the wall as a soft coating, so that the rainwater will clean the facade. The coating needs to be re-applied every ten years. To avoid moisture and dust entering the joints of the glass blocks, they are sealed with a moisture- and water-resistant adhesive of the same UV-modified acrylate family, yet with a significantly higher viscosity. This adhesive is also resistant to glass detergents. Both adhesives are aging resistant and do not discolour when exposed to direct sunlight. A wall specimen has been tested in sunlight over a period of six months and neither discolouring nor penetration of moisture has been observed.

3.1Compression tests

In the primary research stage, four columns were tested in a force controlled hydraulic compression machine, to determine the compression strength of the glass block-adhesive structure as an assembly. The columns comprised standard Poesia blocks of 53 × 116 × 246 mm (specimens A and B) and of 53 × 116 × 121 mm (specimens C and D) that were adhesively bonded together in three different configurations (see Figure 9), to study how the latter affect the strength. Two 18 mm multiplex plates were put at the top and the bottom of each column to prevent the direct contact of glass with metal.

To verify the high compression strength of the final glass blocks, three series of blocks of the custom-made sizes (65 × 210 × 105 mm, 65 × 210 × 157.5 mm and 65 × 210 × 210 mm) were tested in a displacement controlled hydraulic compression machine. The maximum load capacity of the machine was 3 MN. In the first two series of tests, the blocks were placed directly on the machine’s metal surface, while in the third series, two 18 mm multiplex plates were placed between each glass block and the steel surfaces of the compression testing machine. For safety reasons, all specimens were wrapped in several layers of clear PET plastic foil and put in a steel safety cage with polycarbonate windows.

3.2Four-point bending tests

To determine the flexural strength of the glass masonry system, three prototypes were tested in four-point bending until failure in a Zwick Z100 displacement controlled universal testing machine, using a specially fabricated test frame.

All prototypes were made using annealed Poesia solid glass bricks bonded together into a beam by the selected UV-curing adhesive. The specimens had different dimensions and configurations. The initial prototypes, A and B were made using standard sized Poesia bricks (53 × 246 × 116 mm). The last prototype C was made using the final-sized bricks (65 × 210 × 210 mm). To determine both the in- and out-of-plane bending strength of the wall, the specimens were placed either standing or lying in the test frame. The dimensions, configuration and experimental set up of each specimen are summarized in Table 3.

Prior to testing, all specimens were wrapped in several layers of clear PET plastic foil as a safety precaution.

3.3Impact and vandalism test

The glass masonry facade may in actual usage be subject to impact from a variety of causes, such as the accidental impact from bicycles, skateboards, etc., or to the sustained attack with objects such as bottles, bricks, tools etc. in the case of vandalism. Therefore, a rigid body impact test and a vandalism test were performed on an experimental glass wall. The mock-up consisted of 22 glass blocks, each measuring 53 × 246 × 116 mm, which were adhesively bonded to form a wall (see Figure 10). The glass wall mock-up was mounted into a wooden frame, which was fixed to a rigid concrete wall to simulate the inertia conditions of the glass facade. The specimen was not pre-loaded in compression because such an experiment is difficult to set up in a realistic manner and would not produce a reliable result.

First, a hard body impact test was conducted on the specimen. The test was carried out using a solid concrete brick of 65 × 102.5 × 215 mm, weighing 3.4 kg. The concrete brick was placed in front of the facade, touching the target brick. At that position it was suspended with a hook from a 1.5 m long metal wire, hanging down from a wooden cantilever projecting above the mock-up (see Figure 10). The concrete brick, attached to the wire, was then swung outwards by a 45 degrees angle and released from there. The test was repeated two times from a 45 degrees angle, then another two from 90 degrees angle. Afterwards, a vandalism test was carried out on the same experimental wall using a 4 kg sledgehammer until fracture.

3.4Thermal shock test

On a warm, sunny day the glass blocks can heat up significantly. In the event of rain on the same day, the warmed glass blocks will come into contact with the colder rainwater and limited thermal shock can occur. The shock intensity is related to the temperature difference between the material and the environment and the rate of heat flow from the glass. In that context, a hot-cold thermal shock is more harmful to glass than a cold-hot thermal shock, because it generates tensile stresses on the rapidly cooled surface. These stresses may be sufficient to activate pre-existing micro-cracks and lead to fracture. Therefore, to evaluate the performance of the glass blocks under peak temperature fluctuations, specimens were heated for four hours in a furnace with a constant temperature of 1) 80°C and 2) 60°C. Then they were cooled down by being immediately immersed into water of 20°C for approximately 10 minutes each. Specimens were (A) half-immersed into water, (B) completely immersed into water, (C) immersed only with one face into the water or (D) splashed on one face. The test set-ups are illustrated in Figure 11.

4.1Compression tests

The compression tests of the blocks and columns show that the compression strength of individual glass blocks and of the masonry wall as one structural unit are sufficient to meet the requirements of this design, although the compression strength values of the glass blocks are lower than the ones stated in literature. This can be attributed to the increased amount of interior defects, such as air bubbles, which occur in the casting of such thickness of glass. In a float glass production line, these inconsistencies are avoided through automatic control, but in hand-cast glass they are inevitable. The resulting nominal compressive stresses are summarized in Tables 4 and 5.

The results of the two first series (A, B) of compression tests of bricks, where no intermediary was used between glass blocks and the steel cap of the machine, presented obvious cracks in a compressive stress between 20– 30 MPa, significantly less than the value stated in literature. This is attributed to the high concentrated contact pressure between the stiff glass blocks and the stiffer steel plates of the compression machine (see Figure 12). Any unevenness in the contact surface of the two hard materials induces local peak stresses, which in a brittle material like glass propagate local cracks. This shows the critical importance of properly supporting the glass along the whole surface and not creating any stress concentration in the supports. Therefore, in the third series of compression tests, two 18 mm thick multiplex plates were used as intermediaries on top and bottom of each glass block to ensure an even load distribution (see Figure 12). This demonstrates that the glass blocks can tolerate very high compressive loads. More specifically the smallest block, measuring 210 × 105 × 65 mm, presented its first crack at only 2980 kN load, a load five times as big as the complete dead load of the Crystal House facade, while the other two blocks did not crack until the compressive machine reached its force limit of 3000 kN. This series of experiments emphasizes the importance of designing good connections that optimize the load distribution to the glass masonry wall. Poor detailing or execution can result in high local stresses that significantly reduce the overall strength of the glass structure, while connections that provide a uniform load distribution will result in much higher failure loads. The crack patterns in all broken specimens demonstrate the absence of internal residual stresses in the glass blocks. In specific, no secondary crack branching – an effect of internal residual stress – was observed in any of the specimens, even under high compression loads.

The compression tests of the four glass columns (see Figure 13) revealed significant differences in strength. This can be attributed to firstly the creation of indirect local tensile stresses due to the oblong shape of the specimens, secondly to the different configurations of the glass blocks, and thirdly and most significantly to improper bonding. Indeed, prototypes A and B that presented non-homogeneous bonding demonstrated considerably less strength, up to 50% , than prototypes C and D that were homogeneously bonded. In particular, A and B were adhesively bonded on the largest faces of the standard-sized Poesia bricks. These faces present a convex plane of approximately 0.5 mm at the centre. Thus, the bonding layer is relatively thick and inhomogeneous resulting in a weak joint in the middle of each specimen. In contrast, specimens C and D use a different configuration, where the glass blocks are only bonded on their shorter and much more even surfaces. Accordingly, stronger adhesive bonds were formed and the columns showed a much more monolithic behaviour, resulting in higher compressive strength. Although the samples’ size is limited to derive more general results, they indicate that the compressive strength of the structure is very dependent on the configuration of the blocks and the quality of the bonding surfaces.

4.2Four-point bending tests

The primary four-point bending tests (specimens A and B) indicate an in-plane flexural strength of 7.85 MPa and an out-of plane flexural strength of 9.13 MPa at failure (see Table 6). These results can be used as conservative design values, given the fact that the flexural strength of glass itself is higher. The lower values in fact occur because the beam specimens are only bonded horizontally, resulting to stress concentrations on the open vertical joints, which decreases the strength of the specimen. However, in the actual construction, the glass blocks are confined by the boundaries of the structure and therefore the vertical joints of the wall are prevented from opening and the strength is expected to be higher.

The last specimen was made utilizing glass blocks of 210 × 210 × 65 mm, but from a series that presented considerable intolerances in flatness and height, exceeding many times 0.5 mm, which, given the small thickness of the adhesive, resulted in improper bonded joints. This explains its lower flexural strength (3 MPa) and highlights once more the critical role of the accurate dimensioning of the glass blocks to the overall structural performance of the wall.

4.3Impact and vandalism test

The prototype resisted successfully all impact tests without presenting any cracks; while the concrete brick used as impactor was severely damaged. Accordingly, it is expected that the facade can withstand the accidental impact of normal objects such as bikes, bottles etc.

The vandalism test with a sledgehammer resulted in internal cracks to the aimed glass block, but no damage occurred to any of its adjacent blocks. A second, adjacent block was then hit by the sledgehammer, and the same internal cracking pattern appeared (see Figure 14). The results indicate that 1) a rapid impact force only causes local damage, which does not transfer to adjacent bricks and 2) the damaged blocks still maintain a smooth external surface, hence people do not get hurt if they touch them. It should be noted that an unavoidable by-product of the glass facade is that the glass blocks will already be in high compression due to the self-weight of the structure; thus, they are prone to fail under less load. However, the pre-compression of the blocks is not expected to significantly change the results.

The vandalism test emphasizes the significance of allowing for replacement in case a brick is damaged. A procedure of replacing a damaged brick was developed using the same specimen: first the greatest piece of the damaged block is mechanically removed until only small shards attached to the adhesive are left. The adhesive is then locally heated above 120°C with a hot air blower. This is the transition temperature where the adhesive starts to become viscoelastic and softer, allowing for easy mechanical removal of the last glass shards and of the adhesive layer itself, without damaging the adjacent blocks. A new glass block, machined down by 0.1 mm in dimensions to slide easily into the empty slot, can be then inserted (see Figure 15). Adhesive can be injected into the surrounding seams, using a syringe.

4.4Thermal shock tests

The results of the thermal shock tests are summarized in Table 7. Specimens C and D are the closest simulation of the hot facade’s resistance against summer rain, since in the event of rain, only the external surface of the blocks will be exposed to rainwater. In none of these specimens appeared any cracks. However, all specimens that were half or completely immersed into water after being heated to 60°C or 80°C presented considerable cracks in their interior due to the abrupt temperature change between the surface and the core. More specifically, both samples B developed internal cracks in all of their volume, while in specimens A cracks were generated only in the part that was immersed in water (see Figure 16). In that case, a clear, almost horizontal cut marks the waterline. In both A and B samples, the cracks continued to grow significantly after they were removed from the water. The results suggest that the blocks can withstand the elements if applied in an external building wall such as the case study, where they will be susceptible to a rapid temperature change mainly on their external surface. Nonetheless, the blocks may be susceptible to damage in locations with extreme weather conditions. If the concept is to be used in a less moderate climate than Amsterdam, it is recommended to test for thermal shock using appropriate parameters.