Impact of interface materials on side permeation in indirect encapsulation of organic electronics

The rapid development of organic electronics, especially in the area of optoelectronics, has led to the development of flexible and efficient lighting, displays, and photovoltaic devices.^1–5 The success of this technology is, however, limited by the stability and vulnerability of organic electronic materials upon exposure to moisture and oxygen from typical surrounding environments in which they are used. With the use of protective layers like glass and vacuum deposited ultrathin barrier films, moisture ingress into the devices can be significantly reduced.^6–9 However, when these protective layers are attached to organic electronic devices using adhesive layers (indirect encapsulation), a permeation pathway for moisture ingress into the device exists from the edges of the package.¹⁰ While using indirect barrier encapsulations, edge seals are often used along the perimeter of the devices to slow the rate of moisture ingress from the sides. Commonly used edge seal materials include polyisobutylene and epoxies. These edge seal materials are often the weak link in making the devices last for several years as moisture can permeate through them. Therefore, these materials are often incorporated with desiccants to improve their efficiency.¹⁰ To date, much of the attention has been placed on improving these barrier adhesives and edge sealants to limit side permeation rates into the packaged devices, with some showing excellent performance even under damp heat conditions.

When barriers are applied to devices using an indirect encapsulation, the use of an edge seal or adhesive between the device and the barrier leads to the formation of two interfaces. The first interface is between the device and the adhesive material, and the second interface is between the adhesive and the ultrabarrier film. Due to the imperfect nature of the contact interface, defects can form and provide a secondary pathway for side permeation to enter the package other than diffusion through the bulk of the edge sealant or adhesive. Although water is present in the atmosphere in the form of water vapor, condensation of water can take place inside the capillary-connected network. Since the movement of water in such capillaries is governed by surface energies of capillary walls, changing the materials at the interfaces can play a significant role in controlling the permeation rates. Studies conducted on the development of barrier films and adhesives have focused only on their quality in terms of the water vapor transmission rate (WVTR) through the barrier structure. However, the effect of side permeation along the interfaces of encapsulation has not received significant attention. If the capillaries are formed at the interface and have water permeation through them, then the performance of the edge seal is governed by the faster permeation mechanism, which is either through the interface or through the bulk of the sealant material.

In this work, we demonstrate the effect of surface energy of the barrier material in contact with an adhesive on side permeation of water vapor into packaged Ca devices. This involved the use of a hybrid barrier film (Al₂O₃/SiN_x) deposited onto a polyethylene naphthalate (PEN) substrate that was used to package Ca samples with a clear UV activated barrier adhesive. The interface between the outer SiN_x layer and the barrier adhesive was modified by coating SiN_x with a thin layer (ca. 1 nm) of Al₂O₃ or TiO₂. Contact angle measurements show that the use of Al₂O₃ or TiO₂ coatings was able to significantly change the water contact angle of the barrier's surface, which also translated into a large effect on the observed side permeation rate. Finally, using a mathematical model, we compare the rate of water permeation through the capillaries and bulk of the polymer to show the effect of this interfacial modification on permeation rates into packaged samples.

Hybrid barrier films were prepared by first depositing Al₂O₃ followed by SiN_x onto 125 μm-thick PEN substrates. As shown in Fig. 1(a), the Type A barrier structure is prepared by the deposition of 100 cycles (cyc) of Al₂O₃ by plasma enhanced atomic layer deposition (PEALD) on PEN substrates at 100 °C. After the deposition of the PEALD layer, 500 nm of SiN_x was deposited by plasma enhanced chemical vapor deposition (PECVD) at 110 °C using a Unaxis PECVD tool. To modify the surface energy at the barrier's surface, a 10 cyc of atomic layer deposition (ALD) Al₂O₃ layer was deposited on the outer surface of Type A samples to make the Type B barrier architecture, as shown in Fig. 1(b). Similarly, for Type C barriers, TiO₂ was grown onto the exposed SiN_x layer by the deposition of 3 cyc of ALD Al₂O₃ followed by 7 cyc of ALD TiO₂. Deposition of Al₂O₃ and TiO₂ on SiN_x changes the surface energy of the interface between the barrier and the adhesive. For PEALD depositions, trimethyl aluminum and tetrakis(dimethylamido) titanium were used as metal precursors, and oxygen plasma was used as an oxidizer with plasma power of 300 W for the deposition of Al₂O₃ and TiO₂, respectively, in a Cambridge Fiji Plasma ALD system. The thicknesses of the films were measured on silicon substrates using a Woollam M2000 ellipsometer. The contact angles of the films were measured using Ramé-hart 250 goniometer. During the deposition of ALD films, since the last layer is always exposed to plasma, its surface remains highly activated and immediately measured contact angles are small. Edy et al. have shown saturation in the value of the contact angle of ALD Al₂O₃ films after 2 days of ageing.¹¹ Contact angles were, therefore, measured after storing the samples in the glove box for at least 2 days after fabrication.

Optical Ca tests were used to determine the side permeation rate of moisture into packaged samples. Ca sensors were deposited with dimensions of 5 × 5 mm² and a thickness of 100 nm on a glass substrate cleaned with detergent and solvent. The deposition of these Ca sensors was carried out using an EvoVac thermal evaporator (Angstrom Engineering, Inc.) at a base pressure of 10⁻⁷ Torr with a deposition rate of 2 Å/s connected to a nitrogen glove box. The sensors were deposited in a 4 × 4 array yielding a total of 16 sensors. The barrier films were attached to Ca sensors using a 25 μm-thick, double-sided UV-curable barrier adhesive (Tesa tape) in a glove box under nitrogen. The structures of the different samples are shown in Fig. 1. The list of specific barrier architectures is given in Table I.

For testing, the packaged Ca samples were placed in an environmental chamber at 60 °C/90% RH for 360 h. The samples were taken out of the humidity chamber at 24 h intervals and optical images were taken using an optical scanner with the resolution of 2400 dpi to observe any change in the area of Ca sensors. Details of the scanning method have been previously published.¹² 

The theoretical model describing the impact of contact angles on the diffusion coefficient in capillaries along the interfaces has been developed following the methodology from Yang et al.¹³ Let us first discuss the origin of capillary forces that lead to capillary flow along the interface. Fundamentally, the capillary forces originate from the surface or interfacial energy (γ). In the case of the interaction between a solid body and a liquid media, the surface tension of liquid tries to minimize the total energy of the fluidic system that basically is the origin of capillary forces.¹³ Therefore, the interaction of forces at the trijunction of a liquid, solid, and air interface is described by Young’s law,

where γ_sa is the solid-air interfacial energy, γ_sl is the solid-liquid interfacial energy, γ_la is the liquid-air interfacial energy, and $θc$ is the contact angle between solid and liquid surfaces, as shown in Fig. 2. When the liquid is present between two parallel solid surfaces, a capillary is created as shown in Fig. 3. The velocity of the capillary meniscus in this capillary is governed by surface energies of the liquid and the solid surfaces. In the case of edge seals, the interface materials are different and, therefore, can have different contact angles with the liquid (water).

Figure 3 represents a partially filled capillary of length L, width b, and surface separation h. The contact angles of water with the surfaces of the capillary are θ₁ and θ₂. The total surface energy of the system can be divided into four parts. First is the surface energy of the filled portion of the capillary of length x where water is in contact with both the surfaces. Second is the empty region of the capillary of length L − x. Third is the surface energy of meniscus and fourth is the surface energy at both ends of the capillary along the width b. Since the thickness h of the capillary is extremely small compared to its length and width (for edge seals), all surface energies associated with it can be ignored. Thus, the total surface energy E of the capillary can be expressed as

where $γsa$ is the surface energy at the solid-air interface and $γsl1andγsl2$ are the surface energies at both solid-liquid interfaces.

Using Young's law for both the surfaces, Eq. (2) can be written as

The derivative of Eq. (3) with respect to x gives the equivalent capillary force F applied on the fluid column along the x direction,

Thus, pressure drop $ΔPla$ across the meniscus is determined by

As shown in Fig. 3, with the supply of water without any external pressure, the movement of water front is solely driven by the pressure drop across the meniscus. To derive the relation for the movement of meniscus front in the capillary, an incompressible Navier–Stokes equation is used.¹³ The current system is treated as a one-dimensional time-variant fluid field where velocity $u=u(y,t)$ varies along the vertical direction and with time. The instantaneous position x of the capillary meniscus can be described by the following equations:

Continuity equation (conservation of mass),

Momentum equation (conservation of momentum),

where μ and ρ are the viscosity and density of water, respectively,

Equation (8) shows that the driving force for the movement of meniscus is only due to the surface tension between water and capillary surfaces. The velocity of the meniscus is given by Eq. (9), and the boundary and initial conditions (at time t = 0, x = L₀) are given by Eqs. (10) and (11),

where L₀ is the initial length of the capillary filled with water. Assuming the profile of meniscus to be parabolic, the velocity distribution across the thickness of the meniscus can be represented by Eq. (12),

Equation (12) can be written as Eq. (13), where the parabolic term satisfies the initial condition and no-slip boundary conditions with surfaces having different contact angles,

Putting the values from Eqs. (8) and (13) in momentum of Eq. (7), we get

Averaging Eq. (14) through the thickness of the capillary and neglecting the double derivative result in

For an initial condition of $x′(0)=0;x(0)=L0$⁠,

where

and D is the diffusion coefficient¹³ of fluid in a capillary attached to a liquid source, without any effect from gravity.

Ca substrates were packaged with the barrier films having three different architectures, Type A, Type B, and Type C, as shown in Fig. 4. Type A barrier architectures used a contact interface of SiN_x with a water contact angle of 32°; Type B barrier architectures used a contact of Al₂O₃ deposited by atomic layer deposition, having a water contact angle of 73°; and Type C barrier architectures used TiO₂ deposited by atomic layer deposition with a water contact angle of 63°. These layers were deposited as the outer layer of hybrid barrier films on PEN substrates that were sealed over Ca sensors using a UV curable adhesive. After packaging, the samples were exposed to the controlled environmental conditions of 60°C/90% RH. Figure 4(a) shows a photograph of the sample before exposure to the humid environment. Figures 4(b)–4(d) are the images of Ca samples exposed to humid conditions for ∼360 h with the SiN_x interface (Type A), Al₂O₃ interface (Type B), and TiO₂ interface (Type C) barrier architectures, respectively. It can be seen from Fig. 4(a) that the edges of Ca devices are straight and sharp before exposure to humid conditions. It can be noted from Figs. 4(b)–4(d) that the extent of side permeation is different in all the three samples with Type A (SiN_x) having largest displacement of the edges while Type B (Al₂O₃) having the smallest displacement of the edges. After the testing period of ∼360 h, the average reduction in the thickness of metallic Ca in the samples was 7.4, 2.4, and 7.2 nm giving effective WVTRs of 6.88 × 10⁻⁴, 2.23 × 10⁻⁴, and 6.69 × 10⁻⁴ g/m²/day in the vertical direction for Type A, Type B, and Type C samples, respectively. However, the edges of Ca devices, which are on the outer side of the samples, become blunt and have moved toward the inner direction indicating the consumption of a metallic Ca by water permeating in the horizontal direction. Since the movement of the edges of the Ca is in the milimeter (mm) range and the reduction in Ca thickness in the vertical direction is in the nanomater (nm) range, it can be concluded that the permeation rate of water in the horizontal direction is several orders of magnitude higher than the permeation rate in the vertical direction, and therefore, the degradation of metallica Ca is governed mainly by the permeation of water from the sides of the sample. It should also be noted that the edges of all four Ca devices, which are located at the center of each sample, remain intact. Although their thickness has reduced over time by a few nanometers, the water from the sides has not yet reached to these devices at the center of the sample. This also indicates that the degradation of Ca devices is governed by side permeation. Some white spots also appear in the samples after 360 h, which can be attributed to the presence of particle defects in the barrier films as described by Kim et al.¹² However, the extent of degradation caused by the permeation of water through these defects is significantly smaller than that from the sides of the sample. Therefore, in this study, the impact of water permeation in the vertical direction and through particle defects has been neglected.

The side permeation rates for the encapsulated samples were determined by measuring the length of Ca consumed along the outer edge with respect to its initial condition as shown in Fig. 5(a). The Ca devices that are located at the corners of the samples have higher and uneven permeation due to their proximity to two edges of the sample. Therefore, such Ca samples were not considered for measuring the rates of side permeation. However, Ca devices that are located at the center of the sides of the samples have water permeation only from one direction yielding uniform permeation. Therefore, for determining the rates of side permeation, only Ca devices that are located at the center of each side have been used as indicated by device numbers from 1 to 8 in Fig. 5(a). Plots in Figs. 5(b)–5(d) show the distance moved by the outer edges of all eight Ca devices along the sides of the samples. These plots consist of two regions. First, a horizontal part indicating that there is no change in the position of the edge of Ca devices and then a slope indicating movement of respective outer edges. From these plots, lag time is determined by the slope of constant permeation rate, and thereby, effective diffusivity is calculated for each Ca device using the following equation:¹⁴ 

where l is the initial distance of the Ca edge from the outer edge of the sample, t_L is the lag time, and D_eff is the effective diffusivity from the outer edge of the sample. The effective diffusivity was calculated from the slopes in Figs. 5(b)–5(d) for all Ca devices as shown in Fig. 6. As clearly seen from Fig. 6, the Type A sample (SiN_x) has the highest rate of side permeation with an effective diffusivity of 3.14 × 10⁻¹¹ m²/s, whereas the Type B sample (Al₂O₃) has the lowest rate of side permeation with an effective diffusivity of 0.37 × 10⁻¹¹ m²/s. The effective diffusivity of the Type C sample (TiO₂) is the intermediate of two with a value of 1.36 × 10⁻¹¹ m²/s. The only difference in the barrier architectures of Type A samples used for the encapsulation of Ca devices from Type B and Type C samples is the deposition of ten cycles of ALD coating as shown in Fig. 1. The overall thickness of ALD coatings is about 1 nm and has been used to change the surface properties of the barrier material at the interface with the adhesive without significantly changing any other parameter like surface roughness. All other parameters of the samples are kept constant. Thus, an order of magnitude difference between the effective diffusivity of Type A and Type B samples indicates that the material at the interface with the adhesive plays a significant role in the rate of side permeation.

In the packaged device structures, water can permeate through different paths as shown in Fig. 7. The first path is the permeation from the top through the barrier film, called as intrinsic permeation. This kind of permeation would decrease the thickness of the Ca devices uniformly and would not specifically etch the sides of the Ca devices. The second path is the permeation through the sides of the package. Since the thickness of the adhesive layer and the interfaces of the adhesive with the glass substrates are the same for all three samples, their effect on the experiment should be similar. The only difference in the three architectures is the interface material between the adhesive layer and the barrier. In the insets of Fig. 5, contact angles for water on various barriers are shown. It is observed from the values for effective diffusivity that the smallest contact angle material (SiN_x) at the interface have the highest value of diffusivity (3.14 × 10⁻¹¹ m²/s), whereas the largest contact angle material (Al₂O₃) at the interface has the lowest value of diffusivity (0.37 × 10⁻¹¹ m²/s). These data clearly indicate that the contact angle of the materials at the interfaces can significantly impact side permeation. Thus, in cases where the side permeation through the barrier adhesive is small, the interface between the barrier film and the barrier adhesive can potentially still play a role during the ingress of moisture into the package.

Using the theoretical model described in the Modeling section, a parametric study has been conducted to determine the influence of changes in the surface properties of one of the sides of a capillary, provided all other parameters are kept constant in Eq. (17). Due to experimental limitations, it is not possible to take the cross section image of the interface between the adhesive (polymer) and solid (SiN_x, Al₂O₃, and TiO₂) surfaces using SEM. Therefore, 10 nm height “h” of the capillary has been chosen to demonstrate the significance of change in surface properties. The trend will, however, remain the same for any other value of height based on Eq. (17). Moreover, the Type A sample is the control for the experiment, which consists of SiN_x at the barrier interface with the adhesive. For the fabrication of Type B and Type C samples, ALD layers were deposited on Type A barriers. As ALD films are highly conformal in nature and the number of ALD layers deposited on Type A barriers is only ten cycles, thickness of which is about a nanometer, the surface roughness or morphology is governed by the surface of a Type A barrier. The other surface of the interface is that of the adhesive layer, which is same for all samples. Therefore, “h” has been considered same for all samples.

At room temperature, the surface energy of a water-air interface is 71.99 × 10⁻³ N m⁻¹ and the dynamic viscosity of water is 8.90 × 10⁻⁴ Pa s.¹⁵ The values of diffusion coefficients for a capillary with thickness of 10 nm and different contact angles on both the sides, calculated using Eq. (17), are shown in Fig. 8. It can be seen from the plot in Fig. 8 that the value of diffusion coefficient “D” reduces with an increase in the contact angle. This means that the permeation rate through the capillaries can be reduced by using materials with higher contact angles.

The materials used as edge seal and adhesives are polymeric in nature and can contain desiccant materials.^10,16 Typically, the diffusion coefficient of a polymer material is in the order of 10⁻¹⁰ m²/s or less. Because of better quality of the adhesive material and the presence of desiccants in them, the diffusion coefficient of water vapor through the bulk of the adhesive can be much lower. Therefore, for comparison of diffusion coefficient of water through the bulk of the adhesive with that of the capillaries at the interface, the value of 10⁻¹⁰ m²/s can be used as a conservative upper limit estimate.

From plots in Fig. 8, it is observed that for all values of contact angles (θ₁, θ₂ < 88.5°) of capillary surfaces, the magnitude of diffusion coefficient of water through capillaries is higher than the diffusion coefficient of water vapor through the bulk of the polymer material by at least three orders of magnitude. This shows that the rate of water diffusion through the capillaries can be significantly higher in certain conditions. This can result in the preferential permeation of water through the interfaces due to capillary forces acting between the surfaces of the adhesive and the barrier. As a result, water front through the capillaries along the interface can move ahead to that of the bulk of the adhesive, and then, it can permeate from the capillaries into the bulk of the adhesive leading to an early saturation of the adhesive as compared to that due to permeation from the bulk alone.

However, if the contact angles of both the interface materials increase beyond 88.5°, there is a rapid decrease in the value of diffusion coefficient of water through the capillary and it approaches zero at 90°. Under such conditions, the permeation rate through the capillaries is reduced significantly. When the contact angle reaches 90° and beyond, the side permeation rate is solely governed by the bulk of the adhesive material. This indicates that it would be beneficial to have a material with higher contact angle at the interface.

In this work, we have demonstrated for the first time the impact of interface materials on side permeation in barrier encapsulations. When devices are encapsulated using barriers and adhesives or sealants, due to imperfect nature of the contact, pores or voids may be formed at the interface. Such features can be the result of surface roughness and porous nature of the adhesive material. We have experimentally demonstrated that the surface property of the materials at the interfaces can significantly affect the rates of water permeation, which in turn affects the quality of the packaging used for the device. The data clearly show that the use of surfaces between the barrier and adhesive with high contact angles are beneficial in slowing down side permeation rates when low permeation barrier adhesives and sealants are used for packaging devices.

This research was supported in part by the Center for Organic Photonics and Electronics at Georgia Tech, by the Department of the Navy, Office of Naval Research under Award Nos. N00014-14-1-0580 and N00014-16-1-2520, through the MURI Center for Advanced Organic Photovoltaics (CAOP), by the Air Force Office of Scientific Research through Award No. FA9550-16-1-0168, and by the National Nuclear Security Administration under Award No. DE-NA0002576 through the Consortium for Nonproliferation Enabling Technologies.