Atomic layer deposited Al₂O₃ capping layer effect on environmentally assisted cracking in SiN_x barrier films

Flexible electronics such as organic light emitting devices (OLEDs), printable solar cells, and stretchable conformal sensors are currently under development, especially given their potential low cost and light-weight characteristics. These electronics require packaging technologies that protect them from degradation from environmental factors (e.g., water vapor, oxygen, etc.) while also providing high reliability under mechanical deformation.^1,2 The protection of flexible electronics such as OLEDs and photovoltaics often includes thin-film ultrabarrier technologies that can be produced by a number of vacuum deposited processes. Often, the mechanical reliability of these films is described in terms of metrics such as the crack onset strain (COS) or the number of bending cycles that the film can sustain without showing degradation. These metrics, however, do not account for the details in damage formation and are thus considered insufficient to determine the mechanical reliability of brittle ultrabarrier films on flexible substrates. In general, the fracture mechanics of thin films has been studied including (1) channel cracking, microcracks, viscous flow, warpage, plastic deformation under good adhesion with substrate and (2) delamination and buckling-induced cracking under poor adhesion with the substrate.³ In particular, channel cracking is one of the primary failure modes of brittle films on flexible substrates and thus considered in this study (see Fig. 1).

Plasma enhanced chemical vapor deposition (PECVD) silicon nitride (SiN_x) has been used in ultrabarrier coatings for flexible displays, as it possibly has an effective water vapor transmission rate (WVTR) on the order of 10⁻⁶ g m⁻² day⁻¹.^4–8 Also, the deposition can be completed at low temperatures (∼100 °C) and high rates (∼60 nm/min).^8–11 Environmentally assisted cracking (EAC) in PECVD SiN_x barrier films has previously been quantified by measuring the channel crack growth rate v as a function of driving force for crack extension, G (v-G curve).^12–14 The driving force G can be calculated using the following equation:^3,15

where ε_app and ε_res are the applied and residual strains in the film and E_f^* and h_f are the plane strain elastic modulus and thickness of the SiN_x film. Z represents the dimensionless energy release rate, which is dependent on the elastic mismatch between the film and substrate.¹⁶ The subcritical crack growth rate v was measured at different applied strains and environmental conditions in the v-G curve. Furthermore, polymer substrate relaxation was found to be negligible, increasing driving force less than 1% at COS.¹² However, a strong impact of environmental exposure from water vapor was seen on the growth of cracks in the SiN_x films.^12,13 This previous study revealed that, as a result of EAC in the SiN_x films, extensive channel cracks can be developed at strain levels 35% below the COS.¹² Thus, EAC is central to the failure of ultrabarrier films exposed to mechanical deformation. This leaves the question as to whether or not the fracture behavior of PECVD SiN_x films can be improved through the use of methods that may limit its exposure to moisture and oxygen.

This paper focuses on a specific issue: the impact of an Al₂O₃ capping layer produced by atomic layer deposition (ALD) and its thickness on the fracture and EAC behaviors of PECVD SiN_x ultrabarrier films. The purpose of the capping layer is to limit the exposure of the underlying SiN_x to oxygen and moisture. Although ALD Al₂O₃ has been shown to degrade in moist conditions,¹⁷ its immunity to EAC (as demonstrated below) makes it a good material to investigate its effect as a capping layer on the EAC of the underlying SiN_x. The ALD Al₂O₃ investigated thicknesses, 2 and 10 nm, are more than one order of magnitude less than the 250-nm SiN_x layer such that their COS on polyethylene terephthalate (PET) is higher than that of the SiN_x layer (the COS of 10-nm-thick Al₂O₃ on PET is 1.48% while the COS of 250-nm-thick SiN_x on PET is 0.95%). However, as will be shown below, the effect of the capping layer, having larger COS, does not necessarily mean an increase in the COS of the bilayer film (e.g., the composite Al₂O₃/SiN_x film).

To create the samples, heat-stabilized 125 μm thick polyethylene terephthalate (PET) (Dupont Teijin Films Melinex ST505) substrates were laser cut prior to deposition to a size of 5 mm by 50 mm. Then, 250 nm PECVD SiN_x films were deposited on the PET at a temperature of 110 °C, a pressure of 1 Torr, 20 W rf plasma, and a rate of 10 nm/min. ALD films were deposited in a Cambridge Fiji Plasma ALD system from Cambridge Nanotech Inc. Trimethyl aluminum (TMA) and oxygen plasma were used as precursors for Al₂O₃ with nitrogen (N₂) as the purge gas. In each deposition cycle, the pulse and purge times for TMA were 0.06 s and 60 s, respectively, and oxygen plasma was used for 20 s with the plasma power of 300 W. The N₂ flow rate was maintained at 20 sccm throughout the deposition. The growth rate for Al₂O₃ was ∼1.2 Å per cycle. The deposition was performed at 100 °C, for two thicknesses of 2 nm and 10 nm.

In situ optical and laser confocal microscopy (Olympus LEXT, OLS4100) was used to observe the subcritical crack growth on the surface of PECVD SiN_x film with and without the ALD Al₂O₃ capping layer using a microtensile testing stage (Linkam Scientific Instruments, TST350). We applied the external-load assisted technique¹³ for channel crack growth rate measurement by pulling the samples to a strain of 0.75% (to nucleate cracks), followed by strain reduction steps (during which the velocity of the cracks was measured) to obtain a full v-G curve. The threshold driving force, G_th, for EAC was obtained by determining the strain at which no crack extension (within the ∼1 μm detection resolution⁵) could be observed within 30 min, corresponding to v less than 0.55 nm s⁻¹. Tests were performed in a controlled environment consisting of dry nitrogen (moisture content of 2 ppm), as well as laboratory air at room temperature.

To calculate the driving force G of a channel crack in a bi-layer film on PET, i.e., Al₂O₃/SiN_x/PET, the modeling method created by Long and Dunn was reproduced.¹⁸ 2D plane strain linear elastic finite element models were used in ABAQUS 6.13.¹⁹ Identical dimensions of the specimens tested in experiment were used to create the model geometry. Materials were assumed to behave linear elastically, leading to the driving force G being

where y is the coordinate representing the direction of film thickness, δ(y) is the crack opening displacement where the film is opened, and σ(y) is the tensile stress far from the crack of the tip where the film remains intact, containing residual stress and modulus of elasticity for each layer that undergoes cracking. The crack opening displacement δ was extracted in order to numerically integrate the stress and obtain the value of driving force G in Eq. (2). Three different channel cracking modes were numerically investigated (cracking only in the SiN_x layer, cracking only in the Al₂O₃ surface layer, and cracking of the bi-layer). In the case of cracking only in the SiN_x layer, COS was calculated as 1.62% and 1.80% for a 2 nm and 10 nm Al₂O₃ capping layer, respectively. In the case of cracking only in the Al₂O₃ surface layer, COS was calculated as 8.91% and 4.36% for a 2 nm and 10 nm Al₂O₃ capping layer, respectively. These large strain values are due to the presence of the SiN_x underneath (much thicker than Al₂O₃) that leads to a large decrease in Z for this cracking mode. In the case of cracking of the bi-layer, COS was calculated as 0.94% and 0.9% for a 2 nm and 10 nm Al₂O₃ capping layer, respectively, and is lower than the COS of the SiN_x single layer films. Therefore, the channel cracking of the bi-layer corresponds to the lowest COS values and is the expected cracking mode corresponding to the crack observations in Fig. 1.

The experimentally measured mechanical parameters outlined by Herrmann et al.,²⁰ Ylivaara et al.,²¹ and our previous work¹² were used. Modulus of elasticity and Poisson’s ratio were taken to be E_a = 150 GPa, v = 0.23 for Al₂O₃ layer,²⁰ E_f = 123 GPa, v = 0.253 for the SiN_x layer,¹² and E_s = 4.07 GPa, v = 0.3 for the PET substrate.¹² Nanoindentation was used to measure the modulus of the SiN_x film E_f. The uniaxial tensile properties of the PET substrates were obtained using the microtensile stage. The residual stress of Al₂O₃ film is taken to be σ_a = 385 MPa, the summation of intrinsic residual tensile stress, i.e., 550 MPa,^4,21 and the residual compressive stress from the coefficient of thermal expansion (CTE) mismatch, i.e., 165 MPa. CTE values of 4.2 × 10⁻⁶/K for Al₂O₃ films²² and 19 × 10⁻⁶/K for PET substrates were used. The residual stress of the SiN_x film was taken to be σ_f = −185 MPa.¹² Figure 2 shows the effective Z value, Z_eff, calculated based on the numerical G values [Eq. (2)] and using Eq. (1) that only considers the SiN_x layer, as a function of capping layer thickness, with and without including the residual stresses. The subscript “effective” was added since the change in driving force G is directly reflected in Z while other parameters in Eq. (1) remain unchanged. This graph highlights the importance of considering residual stresses for accurately calculating driving force G. More specifically, it shows an increase in Z_eff (and therefore G) of 10% (from 12 to 13.2) when adding the 10-nm Al₂O₃ capping layer and considering residual stresses, whereas it would only be 6% (from 10.4 to 11) if ignoring residual stresses.

Figure 3 shows the experimental measurements of COS [corresponding to the applied strain, ε_app, in Eq. (1) at the onset of channel cracking] as a function of capping layer thickness, h_a. The presence of a 10 nm capping layer results in a 5% decrease in COS (from 0.95% to 0.9%), while there is no noticeable decrease with the 2 nm layer. Figure 3 also shows the predicted COS for the bilayer film, as a function of h_a. These predictions were based on Eq. (1) (with Z_eff shown in Fig. 2) and the calculated fracture energies for SiN_x and Al₂O₃, Γ_f and Γ_a, respectively. Based on a COS of 1.48% for 10 nm Al₂O₃ on the PET, Γ_a was calculated to be 6.4 J/m² (using similar finite element models), corresponding to a fracture toughness (K_Ic) value of 1.01 MPa m^1/2. Γ_f was previously found to be 25.2 J/m², corresponding to a K_Ic value of 1.82 MPa m^1/2.¹² The fracture energy of the bilayer film can be calculated as an average fracture energy,¹⁸ 

The model is consistent with the experimental results, predicting a 5%-6% decrease in COS with the presence of the 10 nm capping layer, which is the combined result of a 10% increase in Z_eff (see Fig. 2) and a 3% decrease in G_c (24.5 J/m² for h_a = 10 nm vs 25.2 J/m² for h_a = 0). This decrease in COS may appear counterintuitive given that the COS of 10 nm Al₂O₃ on PET is much higher (1.48%, i.e., a stronger capping layer). In fact, the actual decrease implies that the presence of the capping layer does not affect the nucleation of channel cracks from pre-existing defects in the SiN_x layer; once a defect grows into a channel crack in the SiN_x, the presence of the capping layer does not prevent its propagation (it actually makes it easier due to the increase in driving force G shown in Fig. 2).

Figure 4 displays the v-G curves for 250-nm-thick SiN_x films on PET at room temperature in air and dry N₂ environments. The difference in the two data sets indicates EAC in the SiN_x films.^12,13 The graph also shows the v-G curves in air for the SiN_x films with the 2 and 10 nm Al₂O₃ capping layers. The error bars of empty and solid symbols represent the standard deviation from average rates calculated over 10 and 5 measured growing cracks, respectively.¹³ Most of the solid data points are within the error bars of the empty square data points. Hence, the capping layer is not very effective in preventing EAC of the underlying SiN_x, even though EAC was not observed on ALD Al₂O₃ deposited on PET. (Specifically, the external-load assisted technique¹³ was applied for 10-nm-thick Al₂O₃ on PET from COS of 1.48% to an applied strain of 1.3% in air, corresponding to G = 5.1 J/m² and thus G/G_c = 0.8. No further growth of channel cracks was observed in the Al₂O₃ films, corresponding to crack growth rates below 0.55 nm/s, while crack growth rates of >50 μm/s were measured in SiN_x on PET at the same value of G/G_c.) One possible explanation is that, since the failure mode is through-thickness cracking of the bilayer, the cracked capping layer does not prevent water and oxygen molecules from reaching the crack tip in SiN_x at large driving forces. Nonetheless, Fig. 4 reveals for the capped SiN_x larger threshold driving force values G_th, below which no EAC is observed within the resolution of the technique (<1 μm crack growth in 30 min, corresponding to v < 0.55 nm/s). For SiN_x alone, G_th = 0.17 G_c (or K_th = 0.41 K_c), whereas for SiN_x with the 2 nm (resp. 10 nm) Al₂O₃ capping layer, G_th = 0.225 G_c (resp. G_th = 0.25 G_c), or equivalently, K_th = 0.47 K_c (resp. K_th = 0.5 K_c). Given that the threshold regime is often associated with steric hindrance of the environmental species, it is possible that the 30%-50% improvement in G_th is related to a beneficial effect of the capping layer in further preventing access of water molecules to the crack tip in SiN_x at these low applied driving forces.

The main reason for the lack of significant effect of the alumina capping layer on reducing EAC in SiN_x is the fact that the dominant failure mode is cracking of the bi-layer. This is confirmed in Fig. 5 showing the lowest predicted COS for this failure mode over the other two possible modes for the whole investigated capping layer thickness regime. In order to have an effective capping layer preventing exposure of SiN_x to the environment, the lowest COS needs to be associated with the cracking of the SiN_x only. Our parametric study showed that this is only the case by considering an organic capping layer with a low modulus (E_a = 1.5 GPa) and high fracture energy (Γ_a = 100 J/m²). With these parameters an increased Z (due to the increase in elastic mismatch) between the capping layer and SiN_x results in decreased COS for cracking only in the SiN_x layer. At the same time, as the thickness of capping layer increases, the high fracture energy of the capping layer increases the average fracture energy. Thus, the COS of the bi-layer increases while COS for cracking only in the SiN_x layer stays unchanged. As a result, cracking only in the SiN_x layer becomes the dominant failure mode at ∼1.35% when the thickness of capping layer, h_a, is larger than 95 nm (see Fig. 6). Note that, given the large Γ_a, cracking of the capping layer only is the dominant failure mode for capping layers thicker than ∼200 μm. Although organic films have typically poor WVTR on the order of 1 g/m²/day^23,24 (e.g., PET substrate used in this work shows 4, 2.9, and 0.86 g/m²/day for 125, 175, and 250 μm thick films at 38 °C/90% rh, respectively), thick (e.g., up to 200 μm) capping layers could be used in reducing EAC in SiN_x by delayed exposure of SiN_x to the environment. Additionally, these thick layers may act as a protective layer from mechanical abrasion for the hard barrier layer underneath.

In this paper, ALD Al₂O₃ capping layers were used on PECVD SiN_x barrier films in order to explore alteration in EAC behavior. While crack growth rates were not greatly reduced, thresholds of driving force G_th were increased, possibly as a result of improved steric hindrance of the environmental species thanks to the capping layer. In a parametric study, the results show that for a high fracture energy, a low elastic modulus capping layer can cause a transition of the failure mode to cracking only in the SiN_x layer. With the help of this capping layer, we can expect protection from EAC degradation during mechanical loading.

This work was supported by the National Science Foundation through Award No. 1400077.