Morphological evolution and migration behavior of silver thin films on flexible substrates during thermal cycle testing

Abstract

To understand morphological evolution and migration behavior caused by stress migration and thermal fatigue of silver thin films on flexible substrates, bending tests were conducted with different bending angles for two types of thermal tests. Furthermore, the effect of stress migration and thermal fatigue on the electrical conductivity of the silver thin films was discussed. The experimental samples were silver thin films printed on flexible paper for these tests. In an ultra-low cycle thermal test, the electrical conductivity of a silver film decreased with increasing bending angle because of the occurrence of voids caused by stress migration. Meanwhile, a high-cycle thermal test caused cracks to form and caused further deterioration of the electrical conductivity owing to a combination of stress migration and thermal fatigue.

Introduction

Flexible electronic devices in which electrically conductive thin films are applied to flexible substrates have found applications in rollable displays, ¹ wearable devices, ² flexible photovoltaic cells, ³ and sensors ⁴ because of their portability, flexibility, and electrical conductivity. Recently, much attention has been paid to the issue of the reliability of flexible electronic devices, in particular with regard to the mechanical properties of the flexible substrates and degradation effects in the coated thin films.^5,6 Bending frequently occurs in flexible devices and causes tensile and compressive stress in the films, which affects their electrical characteristics. Furthermore, owing to continuous demand in flexible electronics for smaller devices with larger power loads, Joule heating is expected to soon become a major concern with regard to reliability. It should be noted that thin films experience repetitive ON/OFF currents, leading to sharp temperature fluctuations equivalent to the sort of thermal cycling caused by Joule heating. Two physical phenomena, stress migration (SM) and thermal fatigue (TF), should be correctly managed to ensure a high device reliability under bending during a thermal cycling process. SM is a phenomenon wherein atoms diffuse from high compressive stress regions toward low compressive stress regions owing to hydrostatic stress gradients. TF is caused by repeated stress that alternates between tensile and compressive stress owing to thermal expansion mismatches between the thin films and the substrates. The combination of SM and TF may cause cracks, voids, and hillocks to form in the thin films, leading to a deterioration of the electrical conductivity.^7,8 TF failure of metallization interconnect lines subjected to alternating currents has been reported,^9–12 and the damage and failure were caused the deformation-induced motion of dislocations to the grain boundaries. However, TF in the printed Ag film has not been investigated. Nowadays, silver is a main component of flexible thin films; therefore, it is important to investigate the effects of SM and TF on the electrical conductivity of silver thin films under bending in thermal cycle tests.

The effects of bending on the reliability of flexible electronic devices has already been reported in the literature. Li and Jin ¹³ reported the appearance of microcracks after cyclic bending tests of indium tin oxide films deposited on a polyethylene terephthalate substrate; the electrical resistance increased nonlinearly with a sufficiently large number of cycles. Billah et al. ¹⁴ reported changes in device performance of flexible amorphous indium gallium zinc oxide thin-film transistors under repeated tensile or compressive bending stress; these changes were found to be related to the generation of oxygen vacancies. However, thus far, there has been no investigation in the literature into the interaction of thermal cycling and bending. Thermal cycling tests have been studied extensively in traditional integrated circuits. Ri et al. ¹⁵ reported the TF of Al thin films under high-cycle thermal testing; hillocks and voids were observed on the surfaces of the stressed Al films. Park et al. ⁹ used alternating currents to generate Joule heating cycles in copper films and investigated the effect of the microstructure on TF damage. However, these thin films were deposited on silicon substrates, which cannot be bent.

We have reported a silver pattern on a flexible paper substrate with a good electrical conductivity that was deposited using a microwave sintering method. ¹³ As an extension of our previous work on the reliability of Ag patterns on flexible substrates, we performed bending at different angles in two types of thermal tests in this study (ultra-low and high-cycle thermal tests), in which the total holding times at a high temperature were the same. It was noted that the electrical conductivity decreased as the bending angle was increased in both thermal cycle tests owing to the appearance of voids and/or cracks caused by atomic diffusion. The deterioration of the electrical conductivity was greater in the high-cycle thermal test than in the ultra-low one.

Experimental procedures and results

The sample structure used for testing was similar to that described in our previous paper. ¹⁶ A 300-μm-thick photographic sheet of paper (KA450SCKR, Epson, Japan) was used as the substrate. Then, the conductive Ag ink was dispensed onto the substrate using a writing system. A hot-pressure sintering method was used to improve the conductivity of the written pattern, and the thickness of the Ag film was controlled to be 20 μm (see details in Tang et al. ¹⁶ ).

The experimental conditions are summarized in Table 1. A schematic of the thermal cycle testing apparatus is shown in Figure 1; the apparatus included a linear stage, a ceramic heater, a temperature controller, and a Cu heatsink. The samples were bent at various angles; the sample shown (as an example) was bent at an angle of 90°.

Table 1.

List of experimental conditions and results for each sample.

Sample	Thermal cycle testing type	Cyclenumber	Temperature (°C)		Hold time at hightemperature (s)	Bendingangle (°)
			Lowtemperature	Hightemperature
L1	Ultra-low cycle	1	30	160	20,000	0
L2	Ultra-low cycle	1	30	160	20,000	45
L3	Ultra-low cycle	1	30	160	20,000	90
L4	Ultra-low cycle	1	30	160	20,000	135
H1	High cycle	5000	30	160	4	0
H2	High cycle	5000	30	160	4	45
H3	High cycle	5000	30	160	4	90
H4	High cycle	5000	30	160	4	135

Figure 1.

Schematic of the experimental thermal cycle apparatus.

The samples were studied with two types of thermal testing, namely, continuous ultra-low and high-cycle testing, as shown in Figure 2. The ultra-low cycle test (only one cycle) had a holding time of 20,000 s at 160°C, which was the most bearable heating temperature for the substrate. The thermal cyclic test, meanwhile, had 5000 cycles, and each cycle had a holding time of 4 s at 160°C. Therefore, the total holding time at 160°C was same in each type of test.

Figure 2.

Schematic of the two types of thermal tests.

The surfaces of the samples were observed after the thermal cycle tests were completed and were shown to remain unchanged in the unbent areas as shown in Figure 3(a). The bent areas of samples L2, L3, and L4 after ultra-low cycle testing were shown in Figure 3(b)–(d), indicating that hillocks were formed in sample L2, and voids and hillocks were formed in samples L3 and L4. However, the bent areas of samples H1, H2, H3, and H4 after high-cycle testing were shown in Figure 4(a)–(d), indicating that voids and hillocks were formed in samples H1 and H2, and cracks were formed in samples H3 and H4. It was noted that the cracking was enhanced with increasing the cycle number and the bending angle. Changes in the electric resistance were also recorded as shown in Figure 5, demonstrating that the electrical resistance increased as the bending angle was increased in both thermal cycle tests and that the resistance increased significantly in the high-cycle thermal test.

Figure 3.

FE-SEM images of samples L1 (a), L2 (b), L3 (c), and L4 (d) after ultra-low cycle testing.

Figure 4.

FE-SEM images of samples H1 (a), H2 (b), H3 (c), and H4 (d) after high-cycle testing.

Figure 5.

The relationship between the bending angle and increase in the electric resistance for different tests.

Discussion

Before discussing the interactions between the thermal cycling and bending, we discuss the stress at different bending angles. The bending behavior of the sample was modeled by three-point bending, and a finite element method was used to evaluate the stress of the silver film when the photographic paper was bent at a specific angle. The bending of the sample was assumed to be elastic, and Young’s modulus of the photographic paper and the silver film was set to 5 and 73.2 GPa, respectively. ¹⁴ The stress was calculated based on the elastic conditions. The silver film suffered significant tensile stress, σ₁, when the sample was bent, which increased as the bending angle was increased, as shown in Figure 6(a) and (b).

Figure 6.

The bending behavior of silver films (a) modeled via a three-point bending test using finite element method. (b) Tensile stress variation under various different bending angles.

When the samples were bent at various angles during the heating process at high temperature, an additional tensile stress, σ₂, was generated in the silver film because of the mismatch in the coefficients of thermal expansion between the photographic paper substrate and the silver film, ¹⁷ which enhanced the atomic migration.

In the present samples, silver films are polycrystalline with anisotropic grains; when polycrystalline metallic thin films are subjected to a macroscopic uniform tensile stress, the stress concentrates at the grain boundaries. As a result, a stress gradient is generated from the grain boundary to the inside of the grain, which becomes the driving force behind the atomic diffusion caused by the SM mechanism.^8,18 This mechanism was consistent with the growth mechanism of silver nanoparticles due to SM. Let us consider the atomic diffusion, that is, the number of Ag atoms passing through unit area in unit time, in a silver thin film due to SM. The absolute value of the atomic flux |J| in Ag film can by derived by Herring ¹⁹ and Korhonen et al. ²⁰

$$\left|\mathrm{J}\right|=[C\mathrm{\Omega }{D}_0/\left(k_{\mathrm{B}}T\right)]\exp [-(Q-\mathrm{\Omega }\sigma )/\left(k_{\mathrm{B}}T\right)]\left|\mathrm{grad}\sigma \right|$$

(1)

where C is the atomic concentration, Ω is the atomic volume, D₀ is the self-diffusion coefficient, k_B is Boltzmann’s constant, T is the absolute temperature, Q is the activation energy, and σ is the hydrostatic stress.

It should be noted that stress gradient gradσ is the driving force for atomic diffusion, and the gradσ increases with increase in the stress σ. In the samples L1–L4, the total stress σ was composed of σ₁ and σ₂, and the stress σ was affected by the bending angle during heating at same temperature. Therefore, the surface of the silver film of sample L1, in which no bending occurred, remained unchanged because σ₂ was not large enough to cause mass atomic migration. The atomic diffusion in the sample L4, in which the bending angle was largest, was more serious than the other three samples. This caused more voids and hillocks in the sample L4, which increased the electric resistances of the silver film.

In high-cycle test, the tensile stress σ₂ changes with the temperature, which will lead to TF damage. ⁷ According to the TF damage, the defects, such as dislocations and vacancies, in each grain gathered at the grain boundaries via dislocation glide as the number of thermal cycle increases. Thus, the defect density at the grain boundary increases as the number of thermal cycle increases leading to a lower activation energy, causing an increase in the grain boundary diffusion coefficient and mobility of the grain boundary. Therefore, the TF damage to the polycrystalline can be considered as the effect on decreasing Q in the SM mechanism. So, in the case of high-cycle test, the number of generated voids in silver film is much larger than that of ultra-low cycle heating test. By the combination of TF damage and bending stress, cracks occurred in samples H3 and H4, and the electrical resistances of the silver films in these samples increased significantly.

Conclusion

In this work, the SM and TF mechanism in silver thin films under thermal cycle testing was investigated. SM damage occurred in an ultra-low thermal cycle test owing to bending and heating, and both SM and TF damage occurred in a high-cycle thermal test. The combination of SM and TF likely enhances the amount of atomic diffusion, thus causing the appearance of cracks, which greatly decreases the electrical conductivity of the silver films.