Water penetration
in adhesives
for glass-glass encapsulants
Measuring water penetration for adhesives and
application methods utilizing optical calcium testing
Education: MSc Physics and Astronomy
Track: Advanced Matter and Energy Physics
Universities: VU & UvA
Author: I.M. van Keulen
Student number UvA: 10991417
First corrector: Dr. E. von Hauff
Second corrector: Prof. D. Iannuzzi
Index
1. ABSTRACT ... 3 2. INTRODUCTION ... 3 2.1. GENERAL INTRODUCTION... 3 2.2. CONTEXT... 4 3. BACKGROUND ... 63.1. ORGANIC PHOTOVOLTAICS (OPV) ... 6
3.1.1. Working principles of OPV... 6
3.1.2. Historical development of OPV ... 6
3.1.3. Device structures of OPV ... 7
3.1.4. Performance parameters ... 8
3.2. DEFINING THE LIFETIME OF OPV ... 10
3.2.1. Degradation mechanisms ... 10
3.2.2. Measuring the lifetime ... 11
3.3. IMPROVING OPV LIFETIME ... 13
3.3.1. Mitigating intrinsic instability ... 13
3.3.2. Mitigating extrinsic instability... 13
3.4. RESEARCH QUESTION ... 14
3.5. INGRESS OF WATER VAPOR ... 15
3.5.1. Measuring the WVTR... 15
4. MATERIALS AND METHODS ... 17
4.1. SAMPLE PREPARATION ... 17
4.1.1. Analysed adhesives ... 17
4.1.2. Application of the adhesives ... 17
4.1.3. Step by step approach for the sample preparation ... 17
4.2. OPTICAL CALCIUM TEST CONFIGURATION ... 18
4.3. IMAGE ANALYSIS ... 19
5. RESULTS AND DISCUSSION ... 19
5.1. LAMBERT-BEER LAW VS. LINEAR RELATION ... 20
5.2. CORROSION OF THE CALCIUM LAYER ... 21
5.2.1. Before the 85/85 cycles ... 21
5.2.2. After the first 85/85 cycle... 22
5.2.3. After all 85/85 cycles ... 23
5.3. DISCUSSION ... 26
5.3.1. Properties of encapsulants ... 26
5.3.2. Results under ambient and 85/85 conditions ... 27
5.3.3. Results of the application method on the edges (A) and the sides (B)... 27
5.4. INTERPRETATION... 28 5.5. EXPERIMENTAL UNCERTAINTIES ... 28 5.6. OUTLOOK ... 28 6. CONCLUSION... 29 7. ACKNOWLEDGEMENT ... 29 8. REFERENCES ... 30
1. Abstract
Organic photovoltaics (OPV) is an emerging photovoltaic technology with unique properties, such as semi-transparency, flexibility and light-weight. The record power conversion efficiency has exceeded 18%, with lifetimes reported ranging from months to several years. Despite these interesting properties, OPV has not yet been commercialized on a large scale. Existing commercial feasibility studies identify low lifetime as the limiting factor. The lifetime of an OPV diminishes due to intrinsic and extrinsic degradation processes, of which the latter can be reduced by an adequate encapsulant. A critical property of an encapsulant is the water vapor transmission rate (WVTR).
This work investigates the WVTR of two application methods (inner and outer side) and four types of adhesives (PIB, Torr Seal, Max Repair and silicone) for glass-glass encapsulation. The WVTR is analysed with the optical calcium test, which is based on the corrosion of calcium, due to the reaction with water vapor. The thickness of the uncorroded calcium layer is measured by its transmission. To accelerate the tests, samples are stored in a climate chamber of 85°C and 85% relative humidity (85/85), in agreement with the ISOS-D-3 damp heat test.
Under ambient conditions, the application method of an adhesive on the outer sides and PIB have shown to be best in preventing water vapor from penetrating the sample. Under 85/85 conditions, both application methods performed equally well and Max Repair outperformed all other adhesives.
The study findings provide insight on which adhesive and application methods are optimal for commercializing glass-glass encapsulated OPVs. Future research should evaluate if the 85/85 test conditions provides similar effect on the adhesives over a short period compared to real operating conditions over a prolonged period.
2. Introduction
2.1. General introduction
On the 22nd of April 2016 the Paris agreement was signed by the United Nations Framework Convention
on Climate Change (UNFCCC). The aim of this agreement is to keep the increase in global average temperature rise below 2°C with a maximum increment of 1.5°C. The strategy to reach this goal contains the 20/20/20 target, which stands for 20% reduction of CO2 emission compared to the level in 1990, 20% increase in energy efficiency and 20% of all consumed energy must come from renewable energy [1]. Focusing on the last point, one should note that the increase in global energy consumption in 2018 nearly doubled the average global growth rate since 2010 [2]. One of the driving pillars for this growth is the increasing demand for heating and cooling systems. Additionally, the demand for cooling systems is expected to increase when the temperature keeps rising. Despite the fact that the production of renewable energy has never been larger, the global CO2 emission rose 1.7% in 2018 and set a new record [2].
Although it is next to impossible to make this consuming world less consuming, it is possible to make our consumption less impactful. In order to do so, the proportion of renewable energy to the global energy demand must be increased. According to the International Energy Agency (IEA), renewable energy generation rose up to 14% of all global generated energy in 2019 [3]. In particular, the electricity generation of renewables has the highest increase over the past decade, resulting in a coverage of over 26% of the global electricity generation in 2019. Renewable energy is predominantly derived from hydropower, wind power and mostly from solar power [3]. Nowadays, the most commonly used photovoltaics (PV) is silicon. However, a substantial amount of research is performed regarding emerging PV, which could offer advantages over conventional silicon. These advantages include a low production cost, low energy expenditure and easy upscale possibilities. Furthermore, emerging PV possess properties such as flexibility, semitransparency and light-weight, which could broaden the field of application.
This thesis focussed on organic photovoltaics (OPV), which has attracted strong attention in the past two decades due to its unique combination of properties (i.e. semi-transparency, flexibility, light-weight and simple manufacturing processes). In addition, record power conversion for OPV efficiencies have exceeded 18% with reported lifetimes as ranging from months to several years [4]. Despite their
potentially beneficial properties, OPV has not been commercialized on a large scale. This is because of OPV’s limiting parameters such as high cost of investment, low efficiency but most of all limited lifetime [5]–[13]. It is this limiting lifetime of OPV on which this work will focus.
2.2. Context
PV technologies are considered to be commercially feasible when they reach grid parity, i.e. the levelized cost of electricity (LCOE) ($/kWh), is equal to (or lower) than the cost of electricity from the grid. The LCOE of a technology is defined by the net cost of the PV module (including installation and maintenance) divided by the net electricity produced over the module lifetime,
𝐿𝐶𝑂𝐸 =𝑇𝑜𝑡𝑎𝑙 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑 𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦𝑇𝑜𝑡𝑎𝑙 𝑐𝑜𝑠𝑡 (1) The LCOE therefore allows for a straightforward comparison of the cost of different electricity generation technologies. However, reliable LCOE estimates for emerging technologies can be challenging, as the module efficiency and lifetime are not well-defined. Together, these parameters have a non-trivial impact on the estimation of total energy produced by the module. More specifically, a well-known issue for assessing the LCOE of immature technologies, such as OPV, is their scaling gap. This scaling gap describes the discrepancy between the power conversion efficiencies (PCE) of laboratory-scale devices measured under standard testing conditions, and those of modules operating in location-specific, potentially highly variable, conditions [14]. Further, the module lifetimes are poorly-defined but definitely shorter than the lifetimes of silicon PV modules, and are therefore a serious bottleneck for commercialization [15]. Therefore, although production and maintenance costs may be known, the overall lack of predictability associated with OPV performance over time makes reliable LCOE calculations very difficult.
The difficulty in estimating the LCOE for OPV has led to a variety of estimates for the threshold efficiency required for OPV to achieve grid parity [5]–[13]. This is depicted in Figure 1, where the values for the record efficiencies (from the NREL efficiency chart, accessed on the 6th of June 2020) for
OPV cells (orange circles) and their scaling gap of 40% (blue circles) are shown, along with the estimated threshold OPV efficiencies to achieve grid parity (grey diamonds) based on LCOE assessments that compare the cost of electricity generated by OPV to the cost of conventional electricity from the grid. For reference, energy cost from the grid has increased 0.2 ¢/kWh in the US, over the timespan 2008 – 2020 represented in the graph [16].
Figure 1. Comparison of record PCEs (orange symbols) for lab OPV cells (from NREL) and predictions on requirements for minimum efficiency (grey symbols) in order to achieve grid parity [5]–[13].
It can be concluded that there is a discrepancy between the required efficiency values to reach grid parity, in which no obvious trend can be distinguished. Moreover, the estimated threshold efficiency lies in five out of eight studies below the NREL record lab efficiency of that year [5]–[13]. This highlights the importance of the dissimilarity between lab and fabrication efficiencies.
The discrepancy of the required efficiency values could arise due to the use of different approaches for estimating the cost projections for the LCOE [11]. In addition, the material costs are calculated differently, depending on the device structure, used organic materials, and the fabrication scale [8]. Another contributing factor to the discrepancy is assumed to be the difficulty in defining the lifetime of emerging PV technologies.
A wide variety of estimates of the required lifetime of OPV to achieve grid parity is stated by commercial feasibility studies as well. This is depicted in Figure 2, where estimates for the required module lifetimes are shown according to year of publication. The size of the circle corresponds to the number of commercial feasibility studies that state the same required lifetime. The exact predicted lifetime (range), in years, is stated in each circle. There is no equivalent of record lifetime, like there is for the record efficiency, because there are no standardized testing protocols to determine this.
Thus, little coherence is present on the estimates of the required efficiency and lifetime of OPV to become commercially feasible. However, more coherence is present on the what is limiting OPV to become commercial on a large scale, namely the low lifetimes inherent to OPV. All studies state that one of the limiting parameters for the commercial feasibility of OPV is its low lifetime [5]–[8], [10]– [13] . Of these studies, the majority states that the lifetime of OPV is its only limiting parameter [6], [7], [10]–[12].
Figure 2. The recent estimates for the required minimum lifetime of OPV, in years, in order to reach grid parity. The x-axis indicates the year of publication. The size of the bubbles corresponds to the number of commercial feasibility studies that
3. Background
3.1. Organic photovoltaics (OPV)
3.1.1. Working principles of OPV
In the active layer of an OPV, incident light is absorbed resulting in the generation of a Coulombic-bound electron and hole pair, an exciton. In order to harvest power from an OPV, the exciton needs to be dissociated into free charges. In a heterojunction OPV, this dissociation occurs at the interface of two different organic semiconductors, the electron donor and the electron acceptor. The donor must possess a high ionization potential and the acceptor must possess a high electron affinity. After the dissociation, the free electron and hole are collected at the cathode and anode, respectively. The basic operating principles of OPV are visualized in Figure 3.
Figure 3. Schematic diagram of operating principles in a type II heterojunction OPV. The absorption of the incident light causes the generation of an exciton in the donor. Subsequently, the electron is transported towards the acceptor and cathode,
whereas the hole is transported towards the anode, resulting in the dissociation of the exciton. Adapted from [18].
3.1.2. Historical development of OPV
One of the first demonstrations of a working OPV device was shown by Tang in 1986. This first generation OPV was a thin film bilayer constructed from copper phthalocyanine as the donor and a perylene tetracarboxylic derivative as the acceptor. Tang was the first to demonstrate the donor and acceptor OPV. This solar cell achieved a PCE of approximately 1% [19]. This record PCE was almost 10 times the PCE (0.15%) of the conventional single layer Al/Polymer/Metal solar cells [20]. This 10 fold increase of PCE arose due to a higher exciton dissociation and charge collection efficiency, which underlined the importance of the donor and acceptor (D:A) interface.
In the second generation OPV, which originated in 1995, the donor and acceptor molecules in the active layer were blended, resulting in the bulk heterojunction (BHJ) [21]. Due to this blending, the interfaces between D:A increased. Consequently, the distance that excitons need to diffuse before they dissociate, was reduced. Specifically, if the length of the phase separation between donor and acceptor molecules is within the same order of magnitude as, or lower than, the diffusion length of excitons, there is a high probability that the excitons will be dissociated at the D:A interfaces. In 2001, an OPV based on a conjugated polymer/methanofullerene blend reached a PCE of 2.5% under illumination of AM1.5, as a result of morphology optimisation [22].
In 2004, Siemens further improved the morphology of the donor by using polycrystalline material, in contrast to the amorphous polymer used in 2001. As a result the OPV achieved a PCE of ~5% [23]. The most common electron acceptors in that time were the fullerene derivatives [18]. These acceptors possessed strong electron acceptance and isotropic electron-transport capacities. In addition, fullerene derivatives positively influence the electron delocalization at the donor-acceptor interfaces [24]. However, fullerenes have low intrinsic light-harvesting capabilities [25], [26]. Therefore, nowadays more research is dedicated to non-fullerene electron acceptors (NF). These NFs are proven to have a higher degree of tunability of their absorption spectra and a more efficient charge separation. As a result, NFs ensure higher electron energy levels, lower voltage losses and higher photocurrents than
fullerene acceptors [25], [26]. Consequently, the current record PCE of BHJ OPV is based on NF and is 18,22% [4]. The learning curve of record OPV efficiencies over time is shown in the Best Research-Cell efficiencies chart of NREL in Figure 4 [27].
Figure 4. Top: Best Research-Cell PCEs of emerging PV until 2020 fabricated by NREL, accessed on the 6th of June 2020.
Bottom: part of the Best Research-Cell PCEs of OPV to highlight the learning curve of OPV from 2001 till 2020 [27]
3.1.3. Device structures of OPV
The device structure of an OPV exists out of multiple layers. In the simplest form, an OPV consists of an active layer sandwiched between two electrodes with different work functions. Due to this difference, a built-in electric field is generated that is directed from the low work function towards the high work function, cathode to anode respectively. In order for light to be absorbed in the active layer, one electrode needs to be optically transparent. Therefore, mostly transparent conducting oxide (TCO) materials are used as electrodes. Of these TCOs, indium tin oxide (ITO) is most used, as it possesses both high optical transparency and high electrical conductivity [28], [29]. Besides, it is possible to make the other electrode semi-transparent as well using nanowires or semi-transparent cathodes [30], [31]. As a result, the entire OPV can be tuned in colour and transparency. Furthermore, electron and hole buffer layers are potentially added to the OPV structure, in between the cathode and anode active layer interfaces. These buffer layers prevent the electrons from reaching the anode and contrarily, the holes from reaching the cathode. Additionally, the anode and cathode are not able to diffuse into the active layer due to the buffer layers [32]–[34].
OPVs are fabricated layer by layer, of which the substrate is either a glass or a plastic sheet. During fabrication of lab-scale OPVs, all layers are typically spin-coated, thermally evaporated or vacuum evaporated sequentially [35]–[37]. Nevertheless, spin-coating is not viable for manufacturing fast high-volume OPVs [38]. Consequently, a variety of other deposition techniques are used for the
layer deposition of large-scale OPVs, such as blade coating, slot-die coating, spray coating, screen printing, inkjet printing, and gravure printing [39]. Since none of these techniques require either vacuum or high temperatures, OPVs present an interesting opportunity for low-cost, large-scale PV manufacturing [39].
Due to layer-by-layer fabrication, placement of the anode and electrode relative to the incoming light - and thus the substrate - are reversible. When the anode is directly deposited on the substrate, it is called the standard geometry, and when the cathode is deposited on the substrate it is called the inverted geometry. Inverted geometry is proven to be equally efficient and more stable than standard geometry OPVs [40]–[42]. A schematic overview of the typical structure of an OPV with standard and inverted geometry and the most commonly used material layers can be seen in Figure 5 and Figure 6, respectively. Both standard and inverted geometry have their layers deposited on ITO-coated glass plates. However, in the standard geometry, ITO functions as the anode, whereas in the inverted geometry, it functions as the cathode, provided that ITO is functionalized with a buffer layer.
Figure 5. Schematic overview of the typical structure of a standard geometry OPV, with left: the layer descriptions and right: the most commonly used materials for layers. Adopted from [43]
Figure 6. Schematic overview of the typical structure of an inverted geometry OPV, with left: the layer descriptions and right: the most commonly used materials for layers. Adopted from [43]
3.1.4. Performance parameters
In the dark, solar cells behave as a diode: the equivalent circuit diagram is represented in the circle in Figure 7. When a voltage is applied to the diode, a current called the “dark current” (𝐼𝐷) flows through the device. According to the Shockley model, the 𝐼𝐷 of an ideal diode can be calculated according to
𝐼𝐷= 𝐼𝑆 (𝑒 𝑞𝑉
𝑘𝐵𝑇− 1) (2)
where 𝐼𝑆 is the reverse saturation current of the diode, 𝑞 is the elementary charge, 𝑉 is the applied voltage, 𝑘𝐵 is the Boltzmann constant, and 𝑇 is the absolute temperature of the diode [44].
Under illumination of the solar cell, the generated electrons and holes are transported towards the cathode and anode, respectively. This transport of free charges gives rise to the photocurrent (𝐼𝐿), which has the opposite sign as the 𝐼𝐷. The equivalent circuit diagram of a solar cell under illumination is visualized in Figure 8. When a solar cell is subjected to different biases, the built-in electrical field changes. In an OPV the built-in electrical field is the energy difference between the highest occupied molecular orbital (HOMO) of the donor and the lowest unoccupied molecular orbital (LUMO) of the acceptor. When a reverse bias is applied on the solar cell, both 𝐼𝐷 and 𝐼𝐿 are negative. Therefore,
thermally generated minority carriers, on each side of the built-in electrical field, can cross the built-in field, resulting in a small reverse saturation current (𝐼𝑆). However, this 𝐼𝑆 is slightly influenced by the reversed bias, and mostly influenced by temperature. In addition, this current is so small that it can usually be neglected. When a positive bias is applied to the solar cell, three scenarios can be distinguished, depending on the magnitude of the subjected bias. (i) A positive bias is applied with a magnitude lower than the built-in electrical field. Due to this small bias, 𝐼𝐿 is greater than 𝐼𝐷, resulting in a photocurrent (ii) A positive bias is applied with a magnitude equal to the built-in field. This will ensure that 𝐼𝐷 is equal to 𝐼𝐿, resulting in no net current. (iii) A positive bias is applied with a magnitude greater than the built-in field. Due to this large bias, 𝐼𝐷 becomes greater than 𝐼𝐿 resulting in a drift current. The total current through the solar cell is given by [45]
𝐼𝑇𝑜𝑡𝑎𝑙= 𝐼𝑆 (𝑒 𝑞𝑉
𝑘𝐵𝑇− 1) − 𝐼 𝐿
(3)
Moreover, two different conditions are defined, the short circuit and the open circuit conditions. In the short circuit condition, the applied bias is zero. This results in a photocurrent with the greatest magnitude, according to
𝐼𝑇𝑜𝑡𝑎𝑙= −𝐼𝐿 (4)
whereas in the open circuit conditions the current is zero, resulting in the highest voltage, given by 𝑉 =𝑘𝐵𝑇 𝑞 ln ( 𝐼𝐿 𝐼𝑆+ 1) ≈ 𝑘𝐵𝑇 𝑞 ln ( 𝐼𝐿 𝐼𝑆+ 1) (5)
Figure 7. Equivalent circuit diagram of a solar cell under illumination. Encircled part visualizes the equivalent circuit diagram in dark. Source: Figure 1.23 from [43]
The performance of solar cells is typically measured under standard test conditions (STC), which includes an irradiance of 1000 W/m2 at AM 1.5 G solar spectrum and a temperature of 25°C. During these measurements the characteristic current voltage (J-V) curve of the solar cell is obtained, visualized in Figure 8, which is characterized by various parameters:
The short circuit current (Jsc) is defined as the current at which the applied bias is zero. The Jsc describes the number of charge carriers that are collected at the electrodes. Hence, the Jsc is increased by a smaller bandgap, a higher absorption coefficient, smaller phase separations and a higher carrier mobility. The open circuit voltage (Voc) is defined as the voltage at which the current density is zero. The Voc is influenced most by the bandgap, which is approximately the energy difference of HOMO and LUMO in OPVs. Therefore, the Voc can potentially be enhanced by lowering the HOMO of the donor, by raising the LUMO of the acceptor or a combination of those approaches [46], [47].
The maximal power point (mpp) is defined as the point where the product of the photocurrent and the voltage is maximum, resulting in the highest power generation.
The fill factor (FF) is defined as the curvature of the J-V curve, which can be calculated according to 𝐹𝐹 =𝐽𝑚𝑝𝑝𝑉𝑚𝑝𝑝
𝐽𝑠𝑐 𝑉𝑜𝑐 (6)
where 𝐽𝑚𝑝𝑝 and 𝑉𝑚𝑝𝑝 represent the current density and voltage at the maximum power point, respectively. The FF describes the influence of the internal electrical field on the photocurrent and can be quantified with the aid of the series and shunt resistance.
The PCE is the power conversion efficiency of the solar cell and can be calculated by 𝑃𝐶𝐸 =𝑉𝑜𝑐 𝐽𝑠𝑐 𝐹𝐹
𝑃𝑖𝑛 (7)
where 𝑃𝑖𝑛 is defined as the input power, which is equal to 1000 W/m2 assuming the measurements are performed in accordance with STC.
Figure 8. Schematic representation of an illuminated J–V curve. The short‐circuit point (0, Jsc), the open‐circuit point (Voc, 0), and the maximum power point (𝑉𝑚𝑝𝑝, 𝐽𝑚𝑝𝑝) are visualized as well. The blue dot corresponds to the maximum power point.
Adopted from [48]
3.2. Defining the lifetime of OPV
3.2.1. Degradation mechanisms
The unstable nature of OPV is caused by multiple degradation processes that occur within the device. These degradation processes are very material dependant and may be interdepend. The seven most important processes are highlighted below [49].
1. The intrusion of oxygen, of which the influence is threefold. First, oxygen can oxidize metal electrodes with a low work function, causing formation of an insulating metal oxide layer between the electrode and buffer layer. Consequently, a transport barrier for free charges is created, which results in a S-shaped J-V curve. Second, photon-oxidation reactions can change the energy levels and charge carrier mobilities of donor and acceptor molecules, resulting in a decrease in absorption. Last, more traps are formed, resulting in a lower FF and Voc.
2. The intrusion of water vapor drives three degradation processes as well. First, water damages low work function electrodes, causing electron defects that allow more water intrusion. Second, like oxygen, water vapor can form metal oxide layers between the electrode and buffer layer, making it harder to extract charges. Last, water vapor within the active layer can transfer some fullerene derivatives, which could decrease the stability of the nanomorphology.
Encapsulation
3. The metastable morphology of the BHJ. Due to a high mobility of organic compounds, donor and acceptor molecules can cluster, resulting in a decrease in the number of D:A interfaces.
4. The diffusion of layer compounds into neighbouring layers. As a result, energy levels are changed and charge recombination traps are generated.
5. Degradation caused by irradiance is divided into photochemical and photophysical degradation. Photochemical degradation occurs mostly due to photo-oxidation reactions, which induces three processes stated at the intrusion of oxygen. Additionally, photophysical degradation drives the accumulation of free carriers in trap sites, which decreases the Voc. 6. Degradation caused by heating mainly induces physical degradation of the active layer, the buffer layers itself and their interface with electrodes. Since thermal energy increases the mobility, the nanomorphology is less stable.
7. Degradation caused by mechanical stress mostly occurs during fabrication. It can initiate three degradation processes in the active layer, buffer layers and their interface. First, linear stress on OPVs induces strain in organic materials, which decreases the efficiency. Second, small mechanical stress can give rise to decohesion, resulting in degradation of active and buffer layers. Last, mechanical stress can cause fractures.
There are multiple ways to categorise these degradation mechanisms. One of them is to classify them into photo, air, thermal and mechanical stability. Consequently, a decrease in these stabilities is caused by irradiance, oxygen and water vapor ingress, heating and mechanical stress, respectively. These division and their effects on the OPV structure are shown in Figure 9.
Figure 9. Schematic overview of the influence of the photo-, air-, thermal- and mechanical instability on different layers of an OPV with standard geometry. Adapted from [49].
3.2.2. Measuring the lifetime
The lifetime of a PV module is defined using standardized testing consisting of high stress conditions to accelerate module deterioration. For commercial technologies, such as crystalline silicon, the International Electrotechnical Commission (IEC) 61215 protocol is used. This protocol consists of thermal cycling between –40°C and 85°C and a damp heat test at 85°C and 85% relative humidity (RH) [50]. The module lifetime is determined by the T80, i.e. the time it takes for the efficiency to drop to 80% of its initial value. IEC has developed a protocol specifically for thin film PV modules, the IEC 61646 [51]. This protocol arose due to the fact that each thin film PV technology has specific failure mechanisms [51]. However, it has been widely recognized by PV researchers that these protocols are not suitable or sufficient to determine lifetime values for emerging technologies, for which the specific failure mechanisms are not well understood. [52]–[54].
Emerging PV technologies have many material interfaces which may be susceptible to chemical reactions/degradation. In addition, there is an entire library of materials for fabricating OPV devices. There are countless donor and acceptor molecules, and a multitude of possibilities for combining these to form the solar cell active layer. Identifying and mitigating degradation mechanisms is an active area of OPV research, with activities ranging from the design of stable OPV materials to cell and module encapsulation strategies [54]–[59].
Generally, the loss in OPV efficiency over time is characterized by an initial ‘burn-in’ effect, during which the efficiency rapidly drops 10-50% [60]. After burn-in, efficiency decreases roughly
Photo instability
Air instability
Thermal instability
Mechanical instability
Cathode Electron buffer layer
Active Layer Hole buffer layer Anode
linearly over time, i.e. the ‘term’ degradation. The timescales and magnitude of burn-in and long-term degradation depend on cell composition and architecture, as well as testing conditions. As a result, the research community, specifically the International Summit on OPV Stability (ISOS), developed standardized test procedures that are based on the IEC 61646, but that take into account specific issues related to OPV, such as the diversity of degradation mechanisms. These testing procedures have been extended to other emerging PV technologies, such as perovskite PV [53].
The main goal of the ISOS procedures is to standardize testing conditions in order to facilitate comparisons of PV performance between different laboratories around the world, thereby improving the reliability and reproducibility of reported values, as well as identifying specific degradation pathways (which may have a complex interdependence on a variety of stress factors). The ISOS testing procedures are therefore designed to provide insight into the specific stress factors (and co-dependence of these stress factors) that limit device stability via a series of testing protocols that are designed to isolate degradation mechanisms. In other words, the ISOS procedures are not made for assessing the commercial lifetime of PV modules, but rather, the aim is to provide guidelines for successively sophisticated ageing tests that result in a comprehensive collection of parameters that impact lifetime.
The ISOS community has also proposed to account for the burn-in effect with use of the TS80 lifetime. The TS80is defined as the time when the efficiency of the OPV has dropped to 80% of its value at the start of the long-term degradation, thus after the burn-in effect took place [61]. The burn-in effect and different definition of the lifetime (T80 and TS80) are shown in Figure 10.
Figure 10. Schematic evolution of the PCE of an OPV over time, which illustrates the four pairs of necessary parameters to describe the degradation pattern and lifetime. Adopted from [62].
3.2.3. Challenges in measuring the lifetime
ISOS testing protocols, to measure TS80, provide reliability in comparing OPV devices with each other and offer better insight into device stability. There are, however, multiple challenges in predicting realistic lifetimes for OPV devices, of which three are defined by Roesch et al. (2015) [52].
First, Roesch et al. address the challenges of predicting the lifetime of encapsulated devices, since barrier and adhesive material can prevent oxygen and water vapor penetration within a short term. However, in the long term, complex behaviour of oxygen and water vapor ingress through the encapsulation will take place, resulting in rapid degradation. When this is not taken into account, predictions of encapsulated OPV lifetime will be overestimated. Therefore, it is necessary to obtain a detailed understanding of the lag time for ingress of oxygen and water vapor through the encapsulation.
Second, care must be taken when predicted lifetime values of OPV are used for defining commercial feasibility of OPV. One has to bear in mind that after fabrication, OPVs will not be used until bought and installed. It is not possible to simply combine shelf and operational lifetime to determine the application lifetime. Especially when the encapsulation of OPVs is the limiting factor for lifetime, since ingress of oxygen and water vapor will start immediately after fabrication. Moreover, since ingress of gasses is temperature and humidity dependent, the climate in which OPV devices are used is of importance as well.
Last, in addition to climate dependence ingress, it has been shown that the spectral distribution of the illuminated light source strongly influences OPV stability [63]. Contrarily, another study showed that the deviation of OPV stabilities tested in various climates was less in indoor studies [64]. This highlights the fact that there are still a number of open questions regarding the factors influencing OPV stability.
In addition to these challenges, Roesch et al. proposed an extended ISOS-protocol to reach technological maturity and commercial feasibility in an earlier stage. Nonetheless, they state that, besides the proposed extensions, IEC 61646 protocols will be the ultimate qualification of OPV devices. To conclude, it remains a great challenge to make accurate predictions of the application lifetime of OPV devices. This is a major hurdle, since the lifetime is proven to be the limiting factor for its commercial feasibility.
3.3. Improving OPV lifetime
Apart from dividing degradation processes into photo, air, thermal and mechanical instabilities, a classification can be made where they are divided into intrinsic and extrinsic stability. Intrinsic instability is caused by thermal diffusion of device constituents within OPV layers, which is mostly enhanced by heat and light irradiation [65], [66]. Contrarily, extrinsic instability is a result of oxygen and water vapor ingress.
3.3.1. Mitigating intrinsic instability
Intrinsic instability can be mitigated by four factors [49]. First, by changing the material design of the active layer, for which three factors can be altered: the molecular backbone of donor molecules [67]– [69], the molecular side chains of donor molecules [70] and the use of NF acceptors instead of fullerene acceptors [71]. Second, by change the device engineering of the active layer, for which three factors can be altered as well: the use of a blend of donor, acceptor and a third component [72]–[74], the used processing method [75]–[78] and the inverted geometry [79], [80]. Third, buffer layers can be modified to improve stability [81], [82]. Last, electrodes with improved stability can be implemented [83]–[85].
3.3.2. Mitigating extrinsic instability
An encapsulation, as is shown in Figure 9, can partly prevent oxygen and water vapor penetration in OPVs. Consequently, less degradation of OPVs occurs and lifetime is improved from days to years [86]. Especially for outdoor applications encapsulants are of great importance, since OPVs are exposed to multiple stress factors. The encapsulation typically consists of a barrier film that is attached to the OPV with an adhesive.
3.3.2.1. Barrier material
Barrier films can protect OPVs against water vapor, oxygen, scratches and UV-light. In order to prevent water vapor and oxygen from penetrating, barrier films need to possess a low oxygen transmission rate (OTR) and low water vapor transmission rate (WVTR). Under ambient conditions, the OTR and WVTR of barrier films must be around 10−3 cm3 d−1𝑚−2 and 10−6 g d−1𝑚−2, respectively, to protect OPVs [43]. The OTR and WVTR are influenced by the thickness of the film and the chemical and physical characteristics of the material. In addition, barrier films must have a high stability when they are exposed to UV-light, since photo-corrosion must be prevented. In order to fabricate flexible OPVs, typically poly-ethylene terephthalate (PET) sheets or ultrahigh barrier foils are used as both barrier material and substrate. Perfect barrier materials, which have an OTR and WVTR of zero, are glass and certain types of thick metal plates. However, since glass has a high glass transition temperature, another moulding technique or adhesion is needed to make a fully glass encapsulated OPV.
3.3.2.2. Adhesive material
Besides having a low OTR and WVTR, adhesives used for the encapsulation of OPVs must be solvent free, electrically insulating, non-reactive with OPV layers and possess good adhesion to substrate and barrier film. In addition, it is preferred that adhesives are processed at around room temperature. Since preliminary damage can occur due to heating, especially during long reaction times, the maximum
temperature that can be applied is around 100°C [87]. Moreover, it is needed to prevent shrinkage of the adhesive material, since it can cause stress between adhesive and substrate, which can result in delamination [87].
There are multiple methods of curing adhesive materials to barrier materials. First, it is possible to apply adhesives with atomic layer deposition (ALD). This is used in organic light emitting diodes (OLED) fabrication, since it is perfect for flexible and organic barrier material [88]–[90]. However, this application method is highly expensive [91]. Second, it is possible to make use of UV curing adhesives, which is used for OLED fabrication as well. Nonetheless, this method is expensive and needs to be executed with great care since OPV layers are sensitive to high irradiation exposure [92]. Third, thermally curing of adhesives is used, where heat is applied to soften or harden adhesives [91]. However, this process is limited to a temperature below 100°C [87]. Last, pressure curing can be used, where sufficient pressure for a specific time is applied to the outside of the barrier material to bond to the adhesive [87].
When flexibility is not desired, four types of adhesives are commonly used. (i) Epoxy and acrylic resins, which are used on glass substrates due to their high bonding to glass and their low OTR. To reduce shrinkage of resins, fillers or thickeners, such as silica, can be added. Nevertheless, shrinkage can still occur during curing of the adhesive. Acrylate mixtures possess a high resistance to UV radiation and a variety of chemicals. However, their bond strengths depend on the polarity of their substrates. On polar substrates, such as glass and metals, the bond strength of acrylates is high, whereas it is low on non-polar substrates, such as polyethylene (PE) and polypropylene (PP) [87]. (ii) Pressure-sensitive adhesive tapes, which ensure no shrinking. However, pressure-sensitive adhesive tapes can only cope with a low roughness of the surface of the substrate, resulting in an increased permeation at the interfaces [87]. (iii) Ethylene-vinyl acetate (EVA), which is already used in the fabrication of CIGs [92] and has a WVTR around 40 g/m2/day [91]. In addition, it is able to resist heavy weather and to ensure long term reliability under light exposure [91]. (iv) Butyl rubbers, such as polyisobutylene (PIB), which is used as a spacer in double-glazed windowpanes and CIGs glass-glass solar cells. PIB possesses a low OTR and a WVTR around 10−2 to 10−3 g d−1𝑚−2[93]. Moreover, it can be processed at room temperature [92].
3.3.2.3. Getter material
In addition to barrier and adhesive material, a getter material (scavenger material) can be added. This getter material is encapsulated along with the OPV and reacts faster with oxygen and water vapor than organic materials. Therefore, it functions as an active barrier until its maximum oxygen or water vapor uptake is reached. This maximum uptake is material specific [94]. In the past, getter materials were used for outgassing of vacuum [95] and in the food packaging industry [96]. Since research in the encapsulation of OPVs has copied a lot from the food packaging industry, getter materials were used in the encapsulation for OPVs, resulting in an enhanced shelf-lifetime [97]. However, most scavengers absorb in the solar spectrum, resulting in a negative impact on the PCE of OPVs [98]. This ensured less research into getter material for OPV encapsulation in the past few years.
3.4. Research question
The lifetime of OPVs is influenced by a wide variety of factors, including degradation due to reactions with oxygen and water vapor. Ingress of oxygen and water vapor can be mitigated with the aid of an adequate encapsulation. Consequently, the quality of encapsulants is an important parameter for lifetime. Therefore, in this research the WVTR of four adhesives for glass-glass OPV encapsulation will be analysed. Since glass has an OTR and WVTR of zero, water vapor can only penetrate through the adhesive and the interface between adhesive and glass. Moreover, two different methods of applying the adhesive on glass plates is investigated as well. Based of those results and the cost of the adhesives, a recommendation is made on which adhesive and application method is best to use for commercializing glass-glass encapsulated OPVs. This thesis sought to answer which adhesive and application method is most cost-efficient to use for preventing water vapor from entering glass-glass encapsulated OPVs?
3.5. Ingress of water vapor
3.5.1. Measuring the WVTR
There are several methods to measure the WVTR, which can be divided into two groups, based on their working principle. The first group measures the number of penetrated water molecules, referred to as MOCON, named after the largest manufacturer that makes WVTR permeation analysers based on this method. The second group is based on the reaction of water with calcium. The MOCON method consists of two chambers separated with the to-be-analysed adhesion or barrier film. One chamber is filled with nitrogen gas, the other with humid gas. Due to partial pressure, humid gas is pushed to diffuse into the nitrogen chamber. In order to do so, it has to penetrate through the adhesion or barrier film. The number of penetrated water molecules is measured in the nitrogen chamber with a detector, for which a coulometric detector is typically used. Hence, the WVTR is defined as the detected number of water molecules, when steady state is reached. A drawback of this MOCON method is that it is not possible to measure the low WVTRs required for OPVs with commercial MOCON instruments [99].
Therefore, in this experiment, it was chosen to measure the WVTR according to the method of the second group. This method is based on the chemical reaction of water vapor with calcium, given by 2𝐻2𝑂 (𝑔)+ 𝐶𝑎(𝑠) → 𝐶𝑎(𝑂𝐻)2(𝑠)+ 𝐻2 (8) The colour and electrical conductance of Ca differs from the one of Ca(OH)2. Ca is metallic and electrically conducting, whereas Ca(OH)2 is transparent and electrically insulating. Therefore, it is possible to analyse the reaction from Ca to Ca(OH)2 by measuring the electrical conductance or the transmission and reflectance of the sample, which are called the electrical and optical calcium test, respectively. The record sensitivity of the optical and the electrical calcium test is 3 10−7 g d−1𝑚−2[88] and 10−5 g d−1𝑚−2, respectively [100]. In addition, to keep the number of interfaces of the adhesives as low as possible, it is preferred that no electrodes are connected to the sample. As a consequence, this research uses the optical calcium test for measuring the WVTR. The microscopical transmission images that visualize the transition from metallic to transparent are shown in Figure 11.
The optical calcium test measures the light transmission through a calcium layer with homogeneous thickness, which is described by the Lambert-Beer law
𝐼 = 𝐼0 𝑒−𝛼 𝜏 (9)
where
𝐼0 is the initial light intensity, 𝐼 is the light intensity after passing the layer, 𝛼 is the absorption coefficient of the calcium layer and 𝜏 is the thickness of the calcium layer. The light intensity can be calculated by the transmission according to𝑇 = 𝐼 𝐼0
(10) leaving Equation 9 to
𝑇 = 𝑒−𝛼 𝜏 (11)
From Equation 11 the uncorroded calcium thickness can be calculated, which can be converted to the amount of water in grams that has reacted with the calcium.
Figure 11. Microscopical transmission images of a 120nm Ca layer on glass over time. The chronological order is from left to right and from top to bottom, with a ten seconds time scale between the images of the upper row and thirty seconds time scales between the other images. The chemical reaction of Ca with water vapor ensures that the Ca layer goes from metallic
towards transparent, which can be seen by the increase of white area compared to the black area.
For the optical calcium test, it has been assumed that the calcium corrosion is homogeneous. However, Klumbies et al. (2013) observed on the microscale a significant ratio of inhomogeneous corrosion of the calcium layer [101]. Some areas where hardly or not corroded, whereas other areas were strongly or completely corroded. Due to this inhomogeneous corrosion, the Lambert-Beer law underestimates the amount of uncorroded calcium and thus overestimates the WVTR. Therefore, it is recommended that data is evaluated by the linear relationship between transmission and amount of calcium left, as well as by the Lambert-Beer law. Therefore, the Lambert-beer law accurately predicts the amount of uncorroded calcium in a homogeneous corrosion model, whereas the linear relationship is more accurate when assuming the inhomogeneous corrosion model. Moreover, the deviation between the linear relationship and the Lambert-Beer law depends on the thickness of the initial calcium layer. When the initial thickness is low, the inhomogeneous corrosion is low, contrarily, when the initial thickness is high, the inhomogeneous corrosion is high [101].
According to Boldrighini et al. (2019), the inhomogeneous corrosion is a result of the water vapor ingress from the sides of the sample [102]. Therefore, they propose to measure both the full calcium layer and the central zone of the calcium layer, which is the inner region of the calcium. When done properly, the degree of the different types of corrosions can be distinguished [102].
4. Materials and methods
4.1. Sample preparation
4.1.1. Analysed adhesives
The adhesives that were addressed in this experiment are PIB, Torr Seal, Max Repair and silicone shown in Table 1. First, PIB was used in this experiment because Kempe et al. (2014) found that PIB outperforms the best low diffusivity encapsulants in preventing water vapor from penetrating [103]. However, it must be kept in mind that the different concentration blends of PIB ensure different WVTRs, resulting in a WVTR range of 0.01-0.001 g d−1𝑚−2[93]. Second, Torr seal® was chosen, due to the fact that it has been widely used in the vacuum industry to prevent leaks in vacuum systems. Additionally, it proved to yield the best results as an adhesive for micro batteries [104]. Third, Max Repair is investigated, since it consists of shape memory polymers and possesses a high resistance for moisture, temperature, UV-radiation and chemicals. In addition, it has 0% shrinkage after curing and it is transparent and cost-effective. Last, silicone is used as an adhesive in this experiment due to its low price of €0.04/gram, purchased at Bison Kit® [105]. Furthermore, silicone has recently been used for large-area flexible photonic crystal film stickers for light trapping applications [106] and as a glue for measuring the WVTR of other films [107]. Therefore, it is assumed that silicone has a low permeability for water vapor and could be cost-effective as an adhesive. The adhesives are summarized in Table 1.
Adhesives PIB Torr Seal Max Repair silicone
Motivation Best known [103] Best adhesive for
micro batteries [104]
High moisture and temperature resistance [108]
Used as glue to test WVTR of thin films [107]
Manufacturer Merck® [109] Kurt J. Lesker
Company® [110]
Bison Kit® [108] Bison Kit® [105]
Cost €1.90/gram [109] €1.69/gram [110] €0.29/gram [108] €0.04/gram [105]
Thermal linear expansion coefficient 2.1 10−4 ℃−1 [111] 3.03 10−5 ℃−1 [110] 2.16 10−4 ℃−1 [112] 3.77 10−4 ℃−1 [113]
Table 1. Relevant parameters of the analysed adhesives.
4.1.2. Application of the adhesives
The adhesives were applied on glass plates using two methods. (i) The adhesive was applied on the inner edges of the bottom glass plate and afterwards the top glass plate was placed on top. (ii) The two glass plates were placed on top of each other and afterwards the adhesives was applied on the outer sides of both glass plates, shown in Figure 12 A and B, respectively.
Figure 12. The structures of the glass-Ca-glass samples with A. the adhesive on the inner edges and B. the adhesive on the outer sides.
4.1.3. Step by step approach for the sample preparation During the optical calcium test the following steps were carried out:
1. 2.5 by 2.5 cm glass plates with an ITO layer were washed according to a rinsing protocol, shown in the appendix.
2. 10 x 15 mm² calcium with a thickness of 70 nm ±2.4 was evaporated with the Inficon® vacuum evaporator in the glovebox.
3. The samples were transferred to the glovebox filled with N2 gas. In addition, the glovebox contained <0.1ppm H2O and ~0.1 ppm O2.
4. The adhesives were put either on the inner edges (A) or on the outer sides (B) of the two glass plates. In total 4 mL per adhesive for both the application methods is used, applicated with the same tools.
In total 18 samples per adhesive were fabricated, of which nine each are applicated according to method on the edges (A) and on the sides (B) shown in Table 2. The samples were named after the first letter of the adhesive, the number of the sample and the application method. Examples of the samples are shown in Figure 13.
Adhesives PIB Torr Seal Max Repair silicone
Method A B A B A B A B
Sample # P1A P1B T1A T1B M1A M1B S1A S1B
P2A P2B T2A T2B M2A M2B S2A S2B P3A P3B T3A T3B M3A M3B S3A S3B P4A P4B T4A T4B M4A M4B S4A S4B P5A P5B T5A T5B M5A M5B S5A S5B P6A P6B T6A T6B M6A M6B S6A S6B P7A P7B T7A T7B M7A M7B S7A S7B P8A P8B T8A T8B M8A M8B S8A S8B P9A P9B T9A T9B M9A M9B S9A S9B
Table 2. Overview of all samples. Named after the first letter of the adhesive, the number of the sample and the application method on the edged (A) and on the sides (B).
Figure 13. Examples of the samples. From left to right, from top to bottom: Torr seal, PIB, Max Repair and silicone. The A and B correspond to the adhesive on the inner edges and on the outer sides, respectively. The numbers correspond to the
sample number of the batch (one to nine).
4.2. Optical calcium test configuration
The test consists of three phases. First, initial images of both the full calcium samples (10 x 15 mm²) and the central zone (5 x 10 mm²) are taken, outside the glovebox (specifics of the glovebox are in 4.1.3.). Second, the samples were stored in a climate chamber of Clive Hurley® of 85% RH and 85 °C (85/85), in line with the ISOS-D-3 damp heat testing [54]. Third, after every cycle of 20 days, the full calcium samples, as well as the central zone of 2 or 3 samples per adhesive and application method are analysed, until 3 cycles are done. In addition, it is assumed that all samples of the same adhesive and applying method will react in the same way on the rise in temperature and RH.
The transmission of the cells is measured via the test setup devised by Boldrighini et al. (2019), visualized in Figure 14 [102]. The used light source is an Intralux® DC-1100. In order to analyse only the light through the calcium samples with the camera, two rectangles of 10 x 15 mm² and 5 x 10 mm² were cut from polypropylene Entegris® sample holders with a laser cutter. The used camera and tripod is a AF-S Nikkor 18-70mm of Nikon®, the settings are set manually, shown in the appendix. The total
test setup is located in an enclosure which blocks all light, to ensure that the measured transmission is not influenced by ambient light.
Figure 14. Schematic cross-sectional figure of the test setup used for measuring the transmission of the calcium sample. Adopted from [102].
4.3. Image analysis
From each image its greyscale values (IGray) were obtained with the aid of open-source Java software ImageJ. The modes of the greyscale values of the images were used as an average over the measured region. The range of the greyscale value from ImageJ is originally from 0 to 255, for a 32-bit image. However, the measured grey value of an uncorroded calcium sample with initial thickness corresponded to 12. In addition, the measured grey value of a completely corroded calcium sample was 200. Therefore, the used corrosion range of the greyscale value was from 12, for no corrosion, to 200, for complete corrosion. The greyscale values were evaluated via the Lambert-Beer law and the linear relationship to obtain the remaining calcium thickness.
The difference between the greyscale value of the full calcium sample and the central zone were calculated. Based on this difference, the degree of occurrence of inhomogeneous corrosion was predicted. When inhomogeneous corrosion does not occur, there is no difference expected. In addition, when inhomogeneous corrosion does occur, its degree of occurrence is expected to be higher when the central zone has a much lower grey value than the full calcium sample. With the predicted degree of occurrence of the inhomogeneous corrosion, either the Lambert-Beer law or the linear relationship is assumed to be more accurate.
5. Results and discussion
The transmission of all the samples is measured according to the test setup in Figure 14. Examples of the images of the transmission measurements of the samples are visualized in Figure 15.
Figure 15. Examples of the transmission measurements of the samples, from left to right more degradation of the calcium occurred, due to water vapor ingress. The yellow dashed rectangles show the boarders of the sample.
5.1. Lambert-Beer law vs. linear relation
The transmission can be evaluated by the intensity of the maximum and minimum greyscale value according to
𝑇 = 𝐼𝑚𝑖𝑛 𝐼𝑚𝑎𝑥
(12)
According to the Lambert-Beer law in Equation 11, it follows that,
𝐼𝑚𝑖𝑛= 𝐼𝑚𝑎𝑥∗ 𝑒−𝛼 𝜏 (13) 𝛼 = −1 𝜏∗ Ln 𝐼𝑚𝑖𝑛 𝐼𝑚𝑎𝑥 (14) With the aid of the initial thickness of the calcium of 70 nm (𝜏), the uncorroded greyscale value of 12 (𝐼𝑚𝑖𝑛) and the completely corroded greyscale value of 200 (𝐼𝑚𝑎𝑥), the absorption coefficient of calcium can be calculated by,
𝛼 = − 1 70 ∗ Ln
12
200= 0.040 nm
−1 (15)
This value for the absorption coefficient is in agreement with the literature absorption coefficient of 0.05 ± 0.01 nm−1 [101]. The calculated absorption coefficient of 0.040 nm−1 is used to calculate the Lambert-Beer law between transmission and thickness shown in Figure 16.A. In addition, the linear relationship between the transmission and the thickness is shown in Figure 16.B.
Figure 16. The relationship between the measured transmission and the uncorroded calcium thickness. A, according to the Lambert-Beer law, in which the calculated absorption coefficient of 0.040 𝑛𝑚−1is used. This relationship is accurate when
homogeneous corrosion of calcium occurs. B, according to the linear relationship. This relationship is accurate when inhomogeneous corrosion of calcium occurs.
The differences between the transmission of the full calcium samples and the central zones of all samples, before and after the cycles, is calculated. The probability density function of this difference in transmission is visualized in Figure 17. It can be concluded that there is a higher chance that the central zone is less corroded than the full calcium sample, due to inhomogeneous corrosion. However, the highest probability is that the transmission of the central zone is less than 0.5% lower than the transmission of the full calcium sample. Thus, the difference in transmission between the central zone and the full calcium is extremely small. This is in agreement with Klumbies et al. (2013), who proved that thin layers of calcium, around 60nm, experience little inhomogeneous corrosion [101]. Therefore,
it is expected that the Lambert-Beer law describes the relationship between the transmission and the thickness of the uncorroded calcium more accurately than the linear relationship. Consequently, in the rest of the results the Lambert-Beer law is used to calculate the thickness of the uncorroded calcium according to Equation 14.
Figure 17. The probability density function of the difference in transmission value between the full calcium sample and the central zones. When the difference is zero, no inhomogeneous corrosion occurs, and when the difference is positive, inhomogeneous corrosion occurs. The degree of occurrence corresponds to the difference in greyscale value towards the
positive.
5.2. Corrosion of the calcium layer
5.2.1. Before the 85/85 cycles
During the transfer of the samples from glovebox to test setup, ingress of water vapor already took place. Although transmission images of the samples are taken within two hours, the specific time ranges are not accurate enough to calculate the WVTR. In addition, no steady state of water vapor ingress is present. However, the amount of H2O that has reacted with the calcium can be calculated from the thickness of the uncorroded calcium:
𝑚𝐻20 = 𝑀𝐻20 𝑛𝐻20 (16)
Where 𝑚 corresponds to the mass, 𝑀 to the molecular weight and 𝑛 to the number of molecules or atoms. Since two H2O molecules react with one Ca atom, as can be seen in Equation 7, the equation can be rewritten to,
𝑚𝐻20 = 𝑀𝐻20 2 𝑛𝐶𝑎 (17)
Hence, the number of calcium atoms can be calculated by the calcium mass and molecular weight, according to
𝑚𝐻20 = 𝑀𝐻20 2
𝑚𝐶𝑎
𝑀𝐶𝑎 (18)
Finally, the calcium mass can be calculated by the calcium density and the volume in thickness and surface,
𝑚𝐻20 = 2
𝑀𝐻20 𝑀𝐶𝑎
𝛿𝐶𝑎 𝐴𝐶𝑎 𝜏𝐶𝑎 (19)
Here it is desired to know the mass of H2O per area. Additionally, the molecular weight of water and calcium as well as the density of calcium are known, which leaves the equation to
𝐻2𝑂 𝑖𝑛𝑔𝑟𝑒𝑠 (𝑔 𝑚−2) = 2
18 𝑔 𝑚𝑜𝑙−1
40 𝑔 𝑚𝑜𝑙−1 1.53 𝑔 𝑐𝑚−3 𝜏𝐶𝑎 (20)
𝐻2𝑂 𝑖𝑛𝑔𝑟𝑒𝑠 (𝑔 𝑚−2) = 1.377 106 𝜏𝐶𝑎 (𝑔 𝑚−3) (21) The thickness of the calcium samples is obtained with the aid of the greyscale values, from which the water ingress is calculated, shown in Figure 18. PIB, Torr Seal, Max Repair and silicone are colour-coded in blue, yellow, red and green, respectively. When application method on the edges (A) is used the colour is light, contrarily, when application method on the sides (B) is used, the colour is dark. The error bar shows the standard deviation of the nine samples. In addition, outliers are excluded from the calculation. In the appendix is shown which outliers are present. It becomes clear from the graph that some calcium samples were already exposed to a lot of water vapor. Therefore, the adhesives are divided into two groups based on the degree of corrosion. Group one shows little corrosion, H2O ingress below 0.05 g m−2, and group two shows a lot of corrosion, H
2O ingress above 0.05 g m−2. For the PIB A and the Max Repair A, 3 and 4 samples, respectively, showed little water corrosion, whereas the rest of the samples showed a lot of corrosion. Therefore, the PIB A and Max Repair A samples are divided into little corrosion (+) and a lot of corrosion (-). Besides PIB A+ and Max Repair A+, group 1 contains PIB B, Torr Seal A and Torr Seal B, of which Torr Seal A shows the largest water ingress and PIB B the smallest. Moreover, Besides PIB A- and Max Repair A-, group 2 contains Max Repair B and silicone B, of which Max Repair B shows the smallest water ingress.
Figure 18. The water ingress per adhesive before the samples are exposed to 85/85 conditions. Two groups have been distinguished, based on the amount of water ingress. Left, the group that shows little water ingress. Right the group that
shown a lot of water ingress. PIB A and Max Repair A are divided into two groups, since their samples showed a large deviation. The outliers of the other adhesives are not taken into account.
5.2.2. After the first 85/85 cycle
The WVTR can be calculated by the mass of reacted water molecules, according to
𝑊𝑉𝑇𝑅 (𝑔 𝑑−1 𝑚−2) =𝑚𝐻2𝑂
𝑊𝑉𝑇𝑅 (𝑔 𝑑−1 𝑚−2) = 1.377 106 𝜏𝐶𝑎
𝑡 (𝑔 𝑚 −3)
(23) The samples have been in the 85/85 climate chamber for 20 days. In addition, the thickness of the uncorroded calcium is calculated according to the Lambert-Beer law. The calculated WVTRs per adhesive and application method, according to their colour-code, are shown in Figure 19. The error bar shows the standard deviation of the samples. In addition, outliers are excluded from the calculation. The distinction between the two groups classified on the amount of water ingress before the cycle is maintained. In group one, PIB A and Torr Seal A and B have a WVTR around 0.0034 g d−1 m−2. In addition, PIB B has a slightly higher WVTR of 0.0038 g d−1 m−2 and Max Repair A a lower one of 0.0025 g d−1 m−2. The obtained WVTRs of both PIB A and B, are in agreement with the stated WVTR range from PIB under 85/85 conditions of 0.01-0.001 g d−1 m−2 [101]. However, since PIB A and Max Repair A only contained 3 and 4 samples that showed little corrosion, two uncertainties must be stated. First, PIB A does not have an error bar, due to the fact that only one sample was valid. Second, Max Repair A has large error bar, since there was a great deviation between the two valid samples. Group two, that already showed a lot of corrosion under ambient conditions, has a lower WVTR, due to the fact that there is little calcium left to corrode. Therefore, the WVTRs of group two will not be incorporated into the conclusion.
Figure 19. The WVTR per adhesive after the samples are exposed to 85/85 conditions. The distinction between the two groups based on the amount of water ingress before the cycle is maintained. Left, the group that showed little water ingress
before the cycles. Right the group that showed a lot of water ingress before the cycles. The samples from PIB A and Max Repair A that showed a lot of water ingress and the outliers of the other adhesives are not taken into account.
5.2.3. After all 85/85 cycles
After each cycle, the distributions of the different thicknesses of the samples that are sealed according to the application method on the edges (A) and on the sides (B) are shown in Figure 20 and 21, respectively. PIB, Torr Seal, Max Repair and silicone are colour-coded in blue, yellow, red and green, respectively. The x-axis represents the thickness of the uncorroded calcium thickness and the y-axis the number of samples that have that thickness. One should observe that in Figure 20 the x-axis and the y-axis before the cycles is from 0 to 60 nm and 0 to 4 samples, respectively, while in the other histograms axes and y-axes are from 0 to 9 nm and 0 to 3 samples, respectively. In addition, in Figure 21, the x-axis and the y-x-axis before the cycles is from 0 to 70 nm and 0 to 5 samples, respectively, while in the other histograms x-axes and y-axes are from 0 to 10 nm and 0 to 3 samples, respectively. Moreover, the bins in the histograms are stacked on top of each other to improve data visualization.
From the distribution of the uncorroded calcium thicknesses before the cycles of the application method on the edges (A), the following can be concluded. First, it becomes clear that Torr Seal prevents water ingress under ambient conditions best, with only one outlier. Second, PIB and Max Repair samples can be best divided into two groups classified on the uncorroded calcium thickness. Moreover, the PIB
and Max Repair group with low water ingress contain 3 and 4 samples, respectively. Last, none of the silicone samples contained uncorroded calcium anymore. Therefore, they are excluded from the experiment.
From the distribution of the uncorroded calcium thicknesses before the cycles the application method on the sides (B), a clearer trend can be obtained. From the least to the most water ingress, the order of adhesives is PIB, Torr Seal, Max Repair and silicone. In addition, the thicknesses of the calcium of PIB, Torr Seal and silicone are higher when the samples are sealed with the application method on the sides than with the application method on the edges. Concerning the Max Repair samples, the thicknesses of the application method on the sides fall in between the thicknesses of the two groups of application method on the edges.
After cycle 1, for both application methods, the order is completely opposite. Max Repair samples show the highest uncorroded calcium thickness, followed by Torr Seal samples. In addition, the thickness of PIB is for both application methods the least. Striking is the high thickness of silicone samples of the application method on the sides.
After cycle 2, for both application methods, again Max Repair samples contain the highest uncorroded calcium thickness. Torr Seal and PIB samples have approximately equally low uncorroded calcium thicknesses. Furthermore, no significant difference in thickness between the application methods is present. Moreover, since only four silicone B samples contained a sufficient thickness of uncorroded calcium before the cycles. Therefore, two are put in the first and two in the third cycle.
After cycle 3, for the samples sealed with the application method on the edges, Max Repair ensures again the highest uncorroded calcium thicknesses. In addition, Torr Seal contains a slightly lower uncorroded calcium thickness and PIB contains the lowest thickness. For the samples sealed with the application method on the sides, a different order of thicknesses is present, of which the upper limit is below 5 nm. Two out of three Torr Seal samples ensure the highest uncorroded calcium thickness, followed by Max Repair. Two out of three PIB samples show a slightly lower uncorroded calcium thickness and silicone has the lowest thickness. Like after cycle 2, no significant difference in thickness between the application methods is present.
Figure 20. The distribution of uncorroded calcium thicknesses of the PIB, Torr Seal and Max Repair adhesives applied according to the application method on the edges (A). Four moments are analysed from left to right, top to bottom; before the cycles, after the
first cycle, after the second cycle and after the third cycle. Note that the x-axis and the y-axis before the cycles is from 0 to 60 nm and 0 to 4 samples, respectively, while in the other histograms x-axes and y-axes are from 0 to 9 nm and 0 to 3 samples.
