Challenges in Assessing the Mechanical Behaviour of Coatings on Architectural Glass

Challenges in Assessing the Mechanical Behaviour of Coatings on Architectural Glass

Steve J. Bull

School of Chemical Engineering and Advanced Materials

Newcastle University

Newcastle upon Tyne

NE1 7RU

e-mail:

Introduction

Glass is widely used for many applications but one of its most important applications is in building, so-called architectural glass. For instance in 2012 about 70% of worldwide glass sales were for building applications. About 40% of the worlds energy demand is in buildings, both in terms of heating and air conditioning and lighting. Suitable glazings can reduce energy demand - keep buildings in cold climates warm, in hot climates cool, whilst allowing light through. For this reason, the demand for energy-efficient glazings is rising.

Glass is relatively opaque in the UV but transparent in the visible and IR and for energy efficiency the IR transparency needs to be reduced without compromising the visible light transmission. Low emissivity or energy efficient glazings are usually based on a thin transparent conducting layer which may be used as a single layer of be present as part of a multilayer stack with surrounding anti-reflection and barrier coatings. The main design factor is the optical performance of the coating but there can be issues with mechanical damage particularly due to transport or storage. In most cases coatings are produced by a supplier and the coated product is delivered from a supplier to a fabricator who cuts and assembles the glass into windows. Coated glass may be rejected if it is damaged in transit and the damage affects its optical performance. Therefore an understanding of the mechanical response of multilayer coatings on glass is essential to reduce losses from damage.

Coatings on architectural glass for energy efficiency are generally of two types [1]:

1) On-line coatings – these are coatings which are produced on the float class line generally by a chemical vapour deposition process at atmospheric pressure. Such coatings are usually a few hundred nanometres thick layer of a conducting oxide, such as doped tin oxide. Due to the CVD process the coating is rough and due to the material selection the coatings are hard and difficult to damage. These coatings can therefore be used on the external facing panels of windows and are widely used in domestic dwellings. However, these coatings only show a moderate optical performance and the energy savings they confer are limited.

2) Off-line coatings – these coatings are generally produced by physical vapour deposition (PVD) processes in a separate production facility from the float glass line. The coatings usually consist of multiple layers with a total thickness on around 100nm deposited on standard size flat glass panels (3mx2m). Typically the active layer is a 10nm thick layer of a metal (Silver in the UK) which is transparent at this thickness or less, surrounded by thin oxide layers such as tin oxide and zinc oxide which act as antireflection coatings. The silver produces a layer of mechanical weakness in the coating stack and often an external protective layer is used to reduce stresses that arise from handling on this layer. Also, there is a need for a barrier layer to separate the antireflection coatings from the glass substrate, These layers are often based on doped titanium oxide or silicon nitride. Despite this, the coatings are much less mechanically robust than their on-line counterparts even if they can be very optically efficient. In fact the energy efficiency performance can be double that of the on-line coatings. For this reason these coated glazings are the material of choice for large, substantially glass fronted commercial buildings. The on-line coatings are placed on the inside of a double glazing unit and once this is assembled the propensity for damage is reduced to virtually zero. The key issue in therefore damage introduced during delivery.

The main damage mechanisms that arise during delivery are illustrated in Figure 1 for both types of coatings. A black deposit is clearly visible on the surface of the on-line coatings which obscures visibility whereas a well-defined but random transit scratch is produced on the off-line coating where the coating is stripped down to the bottom of the silver layer. A clear understanding of the delivery process is therefore necessary to understand how such failures are generated.

Both types of coated glass are delivered as large panels on the back of a lorry. The cut glass panels are loaded, almost vertically onto a steel frame called a stillage (Figure 2). Attempts are made to reduce the amount of glass and coating debris released during cutting but it is not always possible to completely remove the debris. Multiple panels are loaded together so there is glass to coating contact and, in order to reduce contact damage, a layer of interlevant material is sprayed on the surface to separate the glass surfaces. The interlevant is typically PMMA spheres of about 70m diameter (Figure 2) which is bigger than the glass debris size and a uniform distribution of these is required to ensure that contact stresses are minimised. In addition to the interlevant packs of ~10 glass sheets are stressed together and may be separated by cardboard spacers. During the delivery process the glass plates may be agitated by acceleration and decleration on the road, exacerbated by poor road conditions, and will move a short distance with respect to each other. It is this movement which is responsible for transit scratches. However, it is unclear which is the precise mechanism of damage in the transit scratch and whether this is due to the interlevant, glass debris or a combination of the two. This requires simulation of the failure in the laboratory.

Simulation of transit scratches

Experiments have been performed to simulate the damage process in the laboratory by trapping a controlled number of interlevant spheres between a small flat glass slider and a larger off-line coated glass plate. The spheres were carefully sieved to ensure that there was a very small size distribution to ensure that the applied normal load was evenly distrusted on all the spheres [2]. From estimates from the interlevant density and the weight and angle of repose of the glass plates in a typical stillage the load on a single interlevant sphere was about 7mN and this was matched in the laboratory. Experiments were carried out with pure interlevant and in cases where there were some small sharp glass debris particle spread on the surface. These experiments showed that the presence of the glass cullet was not necessary to form transit scratches that matched what was observed from commercial failures.

The mechanism of failure can therefore be determined from these laboratory tests. In service the interlevant sphere is loaded elastically on the surface and it significantly deforms. During sliding damage occurs either in the sphere or coating depending on their properties:

In the case of on-line coatings the rough surface of the CVD coating is critical in dictating performance. The polymer spheres are compressed against the coating and the roughness peaks penetrate into the sphere; this is exacerbated by viscoelastic deformation of the sphere such that the true area of contact increases with time until it is almost the same as the apparent area of contact. During sliding the polymer sphere is badly damaged by the rough coating and a layer of polymer is transferred to the coating surface – this is the origin of the black deposit in Figure 1a.
In the case of the off-line coatings the surface is very much smoother and the failures are generated in the coating during sliding. The lateral movement generates compressive stresses at the leasing edge of the contact and tensile stresses at the rear on the contact. This generates through-thickness cracks at the rear of the contact which are diverted along the interface between the silver layer and its underlying antireflection coating (in this case zinc oxide). The effect is that the coating above the silver layer is dragged forward by the sphere and is stripped from the surface as it moves forward. This produces the transit scratch similar to that seen in Figure 1b – the only difference is that the scratches are straight in the laboratory tests whereas they meander in true transit scratches due to the variation in sliding direction which is allowed in the delivery system.

The adhesion failure in offline coated glass is a significant factor in the rejection of coated glass after delivery and therefore it is important to be able to determine the effect of different coatings designs (materials, thickness, layer stacking order, etc.) and different process routes on performance. This involves modelling the stresses responsible for failure and how these are relaxed by fracture or plasticity. The stresses in the coatings are a combination of residual stresses generated during deposition and stresses generated during sliding of a polymer sphere against the surface which depend on the elastic properties of the coatings and the coefficient of friction between the sphere and the topmost coating layer. The fracture behaviour of the coatings depends on the defect distribution and fracture toughness of the individual layers. If these properties are known it is possible to assess the likelihood of failure using finite element modelling of the sliding contact.

Measurement of key properties of multilayer optical coatings

Residual stress

Measurement of the residual stress in a coating can be achieved by a number of methods including curvature and x-ray diffraction methods [3]. The latter is well established and the sin2 method has been widely used in the assessment of thin hard coatings. However, most of the layers tested here are not crystalline and the x-ray diffraction approach cannot be applied to them. The exception is the ZnO layer beneath the active silver layer which shows some crystallinity, though adopts a cubic structure rather than the more stable hexagonal Wurtzite structure due to constraint from the underlying substrate. A residual stress of around 1GPa is measured at room temperature after deposition which combines the stresses generated during deposition at the growth temperature (around 200oC) and the thermal expansion mismatch stresses generated on subsequent cooling. The temperature of the glass substrate increases during deposition and therefore the thermal expansion mismatch stress contribution to the final measure stress increases as the coating thickness increases [4].

The residual stress in the amorphous coatings is measured by the effect it has on the curvature of the coating/substrate system using the well-known Stoney equation [3]. In this case only the thickness of the coating and the thickness of the substrate and its elastic properties need to be known as well as the change in radius of curvature of the sample before and after coating deposition. Thin glass plates were used as the substrate in order to get measureable changes in curvature when very thin coatings were deposited on them. Typical changes in residual stress as a function of coating thickness are shown in Figure 3 [4]. Changes in thermal expansion mismatch stresses and relaxation of stresses by viscous process in the amorphous layers both contribute to the changes in residual stress measured at room temperature as the coating thickness increases.

Substrate properties

Bulk mechanical properties of glass can easily be measured by conventional mechanical testing approaches and Young’s Modulus and Poisson’s ratio needed for modelling are available. However, in the case of float glass the surface properties which might be needed in modelling of scratch failures are not necessarily the same as these bulk properties. Indeed there are often different properties measured for the opposite sides of a float glass sheet since one was in contact with molten tin during manufacture whilst the other was in contact with air. It is conventional to apply coatings to the air side of the float glass for best adhesion so the properties of this side have to be measuredusing indentation tests to generate surface specific data. Hardness, elastic modulus and fracture toughness can be determined from such indentation tests.However, the properties of the glass in the near surface region can show some variation due to compositional changes (leaching, tin uptake, tempering effects) and this should be taken into consideration if good data is to be extracted for modelling [5].

Coating properties

Elasticity and plasticity

It is not possible to measure the properties of the individual coating layers in an optical coating using conventional mechanical testing since the materials do not exist in a form large enough to make a suitable test piece. Furthermore, their microstructure is often very different from that of a comparable bulk material and the properties measured are therefore very different. For this reason it has become commonplace to use instrumented indentation (nanoindentation) tests to measure such coatings [6]. For thicker on-line coatings (~300nm) it is easily possible to measure the properties of the coating independent of the substrate – as a rule-of-thumb the penetration depth must be less than 10% of the coating thickness and an elastic-plastic indentation must be produced to measure the hardness of the coating only depending on the relative hardness and modulus of the coating and substrate [7]. The limiting depth for determining elastic properties is much lower and the elastic properties of the coating (contact modulus) are usually determined by extrapolation of the variation of contact modulus with contact depth to zero depth as outlined in ISO14577.

However, thinner films are generally used in off-line coatings. Practically, given the sharpness of commercially-available nanoindenter tips it is not possible to measure the hardness of a coating that is less than 200nm thick independent of its substrate and with the best possible tips this is reduced to 50nm which is still much larger than the greatest thickness usually used in solar control coating designs. A composite response from both substrate and coating is therefore measured in most tests with such materials. In some cases the deformation-response from the coating substrate system may be modelled to extract coating properties but in most cases an extrapolation approach similar to that in the ISO standard can be applied if it can be shown to be valid.

There are additional concerns that the properties of the coating may be scale-sensitive and therefore show some variation with thickness. To assess this coatings cab be deposited at a range of thicknesses and their properties measured using nanoindentation with an appropriate modelling or extrapolation approach to ensure that the effect of the substrate is accounted for. For coatings in a multilayer stack it is important to deposit the same layers below the coating as might be used in the intended application so that the microstructure of the coating is the same - the variation in coating microstructure with changes in thickness is much small than the variation produced by changing from one substrate to another.

In general, there is no variation in the hardness of the coatings investigated with thickness as these are predominantly amorphous and there is no microstructural length scale (e.g. grain size) affecting their deformation. The exception is the crystalline ZnO coating which shows a pronounced indentation size effect as its hardness increases as he coating thickness is reduced [4]. In cases where plastic deformation occurs this will need to be taken into consideration (e.g. indentation by a sharp indenter) but this is not an issue in the modelling of transit scratches made by a polymer sphere.

In the case of elastic properties there is a slight reduction in modulus as thickness is reduced in the extrapolated to zero depth data but modelling shows this just represents an increasing contribution from the elastic properties of the substrate in the thin coating tests which cannot be corrected for by the extrapolation approach. In this case a modelling approach can be used to show that the coating elastic properties do not vary with thickness [4, 8]. For FE modelling the elastic properties determined here can be used over a wide coating thickness range without introducing significant errors.

Fracture toughness

All optical coatings in glass are too thin to assess by both conventional mechanical tests and indentation fracture toughness tests widely used to assess bulk ceramics since these require well developed starter cracks or produce final crack sizes that are ~10m in length. Cracks are produced in the nanoindentations testing of thin oxide coatings on glass but they have a different geometry (Figure 4). With sharp indenters or at least indenters where the edges of the indenter remain sharp the main failure mode is radial cracking which follows the edges of the indenter and is caused by bending of the coating around the indenter during loading. These cracks are constrained to lie within the impression and they generally penetration through the whole coating thickness but do not propagate into the glass substrate. In cases where the indenter is less sharp its edges are worn) the main type of cracking is picture frame cracking which forms at the edge of the impression as the coating is bent into the dent formed in the glass. To use these cracks for toughness assessment it has been necessary to measure fracture energies from the energy dissipated during the indentation cycle [9-11]. The area beneath the force displacement curve, the work of indentation, represents the energy dissipated in elastic, plastic and fracture processes. When fracture occurs, either through-thickness cracking or adhesion failure, there can be a discontinuity in the load displacement curve if the fracture event is large enough and the energy of fracture can be estimated by subtracting the work of elastic plastic indentation from the total work of indentation before and after fracture [9]. Other approaches can be used when no discontinuity is observed [10]. These energies can be converted to Gc values using experimentally measured crack areas and into fracture toughness values using the measured coating elastic modulus [9].There is no change in measured toughness as a function of coating thickness for any coating which means that the fracture process is controlled by the defects in the coating rather than its bulk properties. Loading rate [12] and multiple loading cycles [13] also have an effect on the observed fracture behaviour.