Extreme Drop Durability of Sintered Silver Traces Printed With Extrusion and Aerosol Jet Processes

Printed hybrid electronics (PHE) assemblies are receiving significant attention because of the new possibilities they offer for three-dimensional (3D), nonplanar electronic systems. PHEs, like many other electronic systems (assemblies of devices, passive components, sensors, electromechanical components, substrates, and interconnects) sometimes have to operate effectively in demanding conditions like extreme temperatures, severe fluctuations in temperature, high humidity levels, and intense shock/drop/impact conditions. Ensuring the reliability of these systems under such harsh application conditions is a crucial task for effective product development [1,2]. This challenge is particularly complex in PHEs, because of the introduction of novel materials, unconventional 3D curvilinear geometries, new manufacturing processes/defects, and new microstructural features such as microscale and nanoscale porosity [3,4]. The properties and performance of PHEs are known to be highly sensitive to their manufacturing methods, process parameters, and material combinations. This study focuses, in particular, on the behavior of PHE circuitry in severe shock conditions, where elevated strains and strain rates can lead to extreme interactions among the substrate, interconnects, and components. Drop-induced damage and failure are estimated to contribute to 20% of the failures in electronics, underscoring the importance of robust methodologies for design and qualification [5].

Nanoparticle-based sintered silver inks are widely favored for printing conductive traces in PHEs, due to their manufacturability, conductivity, ductility, and resistance to oxidation. While considerable efforts have been invested in characterizing the reliability of PHEs based on silver conductors, through such reliability and accelerated life tests as bending [6–10], thermal cycling [9,11,12], and vibration testing [12–14], drop testing has been less extensively studied [15]. Additionally, sintered silver exhibits high sensitivity to strain rate owing to its porous nature, with its properties significantly influenced by microstructure, printing techniques, ink composition, and process parameters [4,16–19]. This study examines the resulting fracture and fatigue behavior for severe shock conditions exceeding 20,000 g, that generate strain magnitudes and rates that are well in excess of what is commonly targeted in the literature and in JEDEC standards [20–22]. The high-g, short-duration impacts targeted in this study can induce resonance in higher-order modes within the tested assembly, due to the high-frequency energy in the short impulse, necessitating systematic studies to assess the durability of printed traces at higher strain rates [23,24]. Such studies are vital for achieving a comprehensive understanding of the reliability profile of printed silver-based traces, and for evaluating the impact of material microstructure, printing techniques, and process parameters.

This paper starts with introducing the test sample design, instrumentation, and equipment used for high-g drop testing. Subsequently, this paper provides: (i) testing procedure and deformation levels of the drop experiments and (ii) details of the transient finite element (FE) simulation of the test sample and experiment. The results of the experiment and simulations are compared and discussed. The paper aims to provide insight into the reliability of sintered silver traces under severe shock conditions by testing traces printed using extrusion and aerosol jet printing (AJP) techniques at peak acceleration reaching 40,000 g. Additionally, this study illustrates the benefits of using a cantilever form factor for standardized drop durability studies of PHE trace materials and substrate materials.

The test sample materials and fabrication methods, testing methods and instrumentation, and finite element modeling methods are discussed here.

The experimental setup for drop testing of the printed silver traces used an FR4 cantilever beam with copper pads for resistance monitoring and mounting strain gauges. The layout of the printed silver traces on the beam was specifically designed to maximize the strain experienced by the trace—by placing trace segments near the clamped end, where the highest curvature occurs during the drop-induced flexure. The cantilever beam samples are 67 mm long, 25 mm wide, and 1 mm thick, with a free unsupported length of 50 mm while the remaining length is secured under a rigid clamp. Further details regarding the dimensions of the sample and silver trace can be found in Fig. 1.

Two techniques were employed to print the silver traces onto the substrate: extrusion printing and AJP. The extrusion-printed traces were created using silver paste, yielding traces with an estimated thickness ranging between 80 and 100 μm. The traces were then sintered at 160 °C for 1 h at ambient pressure. In the case of AJP traces, silver traces were printed as a seed layer and sintered at 160 °C for 1 h at ambient pressure. A layer of copper was then electroplated on the silver traces, enhancing conductivity to the levels achieved in the extruded traces, instead of resorting to multiple layers or passes of silver. The silver thickness was estimated to be 30–50 μm, with copper plated on the surface of the silver to a similar thickness. A layer of polyimide was subsequently deposited over the copper plating and cured at 160 °C for 1 h, to provide protection against oxidation.

Figure 2 illustrates samples produced using both methods, along with optical microscopy images of the final traces following all processing steps. The extrusion-printed silver paste trace exhibited a broad and uniform deposition, potentially sacrificing print resolution for lower resistance and increased trace durability. Scanning electron microscopy (SEM) was utilized to examine the extruded traces for cracks, revealing none, as depicted in Fig. 3. The AJP traces were not subjected to SEM scanning due to the presence of a polyimide layer hindering imaging. The copper-plated AJP trace displayed a thinner profile with periodic transverse cracks in a “bamboo pattern” along the trace's longitudinal direction, resulting from the shrinkage during the sintering process. The cracking observed in the AJP trace was attributed to its relatively high solvent content, which evaporates during sintering, causing silver trace shrinkage. This issue could potentially be mitigated by utilizing oven sintering or implementing a compliant dielectric layer beneath the silver. Although these cracks may compromise trace durability during impact, the continuity of the cracked silver traces was mitigated with the copper over-plating. For each printing method, two samples were printed, as depicted in Fig. 4.

To facilitate the testing process, a drop tower capable of generating shocks up to 100,000 g of acceleration and 0.04 ms pulse width was employed. The drop tower was equipped with a dual mass shock amplifier (DMSA) accessory to ensure consistent shocks and amplify the shock levels to the aforementioned 100,000 g level. Figure 5 illustrates the tower, DMSA, and their respective acceleration profiles. Acceleration was measured using piezo-electric accelerometer sensors. Strain gauges were affixed to designated pads on the cantilever samples, and wires were soldered onto the probe pads using SAC 305 solder to monitor strain and resistance, respectively, as depicted in Fig. 6. Strain measurements quantify the beam response to the drop acceleration and also serve to validate the finite element model of the sample. Resistance measurement acts as an indicator of trace damage. These measurements were captured using a high-speed data acquisition system operating at a sampling rate of 1 MS/s and a resolution of 16 bits to capture any high-frequency behavior during impact. Figure 7 depicts the clamping and instrumentation setup on the DMSA. Optical microscopy and SEM imaging were employed for failure analysis and sample inspection both before and after testing.

The testing procedure for the four specimens shown in Fig. 4 is shown in Table 1. The maximum stress test condition was set at 40,000 g peak acceleration and 0.13 ms pulse width, see Figs. 8 and 9. The 10,000 and 20,000 g profiles have a 0.15 ms pulse width. The acceleration profile measured during impact was found to be consistent and repeatable when measuring the acceleration of multiple drops as shown in Fig. 10. The main purpose of Table 1 and Fig. 8 is to explain the testing methodology and conditions clearly. One specimen from each sample set was tested under step-stress conditions to assess the reliability under different shock levels by dropping them up to 20 times (or till failure) at 10,000, 20,000, and 40,000 g peak acceleration. The remaining two specimens were only tested at the maximum 40,000 g peak acceleration by dropping them 20 times or until failure. Response magnitude (displacement, acceleration, stress, and strain magnitudes) and frequency bandwidth of the test samples increased with the impact g levels.

Dynamic FE modeling is used to estimate the strains at the trace footprint during the impact and also to predict the strains for multiple stress and boundary conditions. Because of the high strains produced during severe shock events, instrumenting such events is difficult since these strains and accelerations can exceed the limits of the sensors used in the test. For this reason, well-calibrated FE analysis, which is used extensively to study the reliability of multiphysics systems, is an important tool to estimate the stresses and strains that PHEs are subjected to Refs. [5] and [24–27].

Additionally, a global-local modeling approach can be used to study the effects of the local geometry of the test specimen on the local stresses, strains, stress concentrations, and strain rates near the observed failure sites in the trace. This paper presents only the global model of the cantilever specimen, and the local modeling is deferred to a future paper. The global model of the sample was modeled through ABAQUS using direct implicit modeling as a 50 × 25 × 1 mm shell model with four-node S4R shear deformable shell elements, see Fig. 11. The element size was set to 1 mm. The model has 750 elements and 806 nodes. The FR4 orthotropic material properties used in the model are shown in Table 2. The density of FR4 was set to 2000 kg/m³ [24]. The FR4 density and stiffness properties were obtained from previous measurements conducted at CALCE. Rayleigh damping was used to dampen the model response, the Rayleigh damping equation is shown in Eq. (1). Alpha was tuned to 300 using trial and error until the simulation results matched the experimental results. Beta was kept at 0. The reason for this is that alpha has a greater influence on the lower-order modes that are of interest here (As discussed later, drop experiments reveal that only the first two bending resonant modes dominate the drop response). In contrast, beta has a greater influence on higher-order modes [24]. The input-g method was used to model the drop excitation, where the sample was base-excited using acceleration measurements recorded at the cantilever base fixture, during a 40,000 g drop. Figure 11 shows the location at which the cantilever specimen was excited.

During the drop testing experiment, strain and resistance were continuously monitored throughout the impact event and the subsequent dynamic response of the specimen. Some cantilevers were equipped with two strain gauges, as depicted in Fig. 12, to ensure precise alignment of the drop and to verify the dominance of bending modes and the absence of twisting modes in the dynamic response. As shown in Fig. 13, the strains exhibited consistent magnitudes and phases, confirming predominantly bending response and proper alignment of the drop. The drop tower alignment and the clamping alignment are important factors for having consistent response in terms of the magnitude and phase, for prevention of exciting twisting modes in the sample. Moreover, it was observed that the beam vibrated predominantly in its first two bending modes, during drops of up to 40,000 g. This was confirmed by measuring the strain at different drop levels and producing FFT plots of the measurements to find the modal frequencies like shown in Fig. 14. At this acceleration level, the maximum strain approached approximately 10,000 με, as illustrated in Fig. 13. Additionally, the strain rate was determined to be approximately 25 s⁻¹ at 40,000 g. Furthermore, an analysis of the relationship between strain and impact acceleration revealed almost linear correlation within the testing range, as depicted in Fig. 15.

The in situ resistance of the silver traces was measured and monitored in real-time during drops and the at-rest resistance was monitored at the start of the test and periodically after a certain number of drops. The resistance at-rest exhibited a gradual increase in the traces until it reached an open-circuit state, as depicted in Fig. 16. Initially, the change in in situ resistance during impact was minimal in the initial drops of the experiment. However, as crack propagation and growth occurred in the silver trace due to repetitive drops, transient resistance spikes were observed during the drop events and were recorded in the measurements. These spikes eventually increased to transient open circuits during flexure, as shown in Fig. 17.

Figure 18 illustrates the number of drops until failure for each sample, both for step-stressed specimens and for those tested at 40,000 g. Silver traces printed with extrusion method exhibited greater durability compared to AJP silver traces, primarily due to adhesion issues and cracking observed in the AJP silver in as-fabricated specimens before drop testing. These issues are further discussed in Sec. 3.2. Nevertheless, traces produced using both printing methods demonstrated sufficient durability to withstand repetitive drop events before experiencing complete failure.

The failed specimens underwent inspection using optical microscopy and SEM imaging. Additionally, some samples were cross-sectioned to provide a more in-depth analysis of the microstructure and thickness of the traces.

• Extrusion-printed specimens: Examination of the extrusion-printed silver traces after drop testing, revealed multiple cracks propagating perpendicular to the principal flexure direction of the beam, as illustrated in Fig. 19. These cracks account for the resistance increases observed during drops and the sudden spikes during impact, indicative of trace degradation. However, complete trace failure can be attributed to inadequate adhesion to the copper pad. Figure 20 illustrates that the extrusion-printed trace delaminated and separated from the probing copper pad, resulting in an open circuit. In contrast, the adhesion of the extrusion-printed silver to the FR4 substrate was found to be excellent, as evidenced in Fig. 21. Additionally, Fig. 18 shows that the microstructure of the extrusion-printed silver is coarse and contains distributed voiding. Unlike in the AJP traces, no significant preexisting cracking from the fabrication process was observed in the trace. The trace thickness measured just below 100 μm.

• AJP specimens: AJP trace samples were inspected using similar procedures as the extrusion-printed trace samples, but without SEM imaging due to the aforementioned polyimide layer on top of the trace. Degradation and failure of the AJP traces were expected to be faster due to the pre-existing bamboo cracking pattern that occurred during the processing steps, see Figs. 2, 16, and 18. Inspecting the AJP traces using optical microscopy revealed missing sections of the traces due to repetitive impact, as shown in Fig. 22. Sintering shrinkage cracking caused stress concentrations that allowed delamination in segments of the silver trace, which can be seen in the cross-sectional analysis in Fig. 23. It is also noted that the fabrication cracking in the AJP silver was so severe that it caused complete disconnection in some of the traces, even before any drop testing. However, the copper plating step re-established electrical continuity by filling the gaps in the silver with copper. When compared with the extrusion-printed trace, the AJP trace has a much finer microstructure and a lower volume of voids. The thickness of the silver is about 30 μm, and the thickness of the copper is about 18 μm.

The flexural strain history obtained from the FE model is compared to the experimental results to verify the model properties. The model and experiment match well, as shown in Fig. 24. The FFTs of the experiment and simulation strains, shown in Fig. 25, confirm the matching of results in terms of frequency of the two dominant response modes but with slight overdamping. In contrast, the model appears to be underdamped when it comes to the third mode (at approximately 3500 Hz). The frequencies of the two dominant modes in the cantilever beam were found to be 200 and 1300 Hz. Finer tuning of the damping will be attempted for future fatigue modeling.

Longitudinal strain at the trace location near the clamp for a 40,000 g drop is estimated to be about 15,000 με, as seen in Fig. 26. This strain level is beyond the range of the strain gauge so there is no experimental comparison shown in Fig. 23. These results are essential for finding the strains at the trace location, as a starting point for future local modeling of the stress and strain in the trace.

This paper provides insights into the reliability of silver printed traces using extrusion and AJP processes (with additional copper over-plating for the AJP traces), under severe shock conditions up to 40,000 g peak acceleration. Additionally, this study has established a starting point for developing a standardized reliability testing technique for PHE trace and substrate materials under extreme shock/impact conditions and high strain rates, by utilizing a cantilever beam platform for testing to maximize the strain the material is subjected to and to simplify the modeling process.

Silver traces printed using extrusion were found to be more reliable and had better adhesion to the FR4 substrate than the copper-plated AJP traces. The AJP silver traces had a bamboo pattern precracking due to sintering shrinkage. However, the copper over-plating ensured the continuity of the traces. The drop test of the cantilever beam specimen was modeled using FE modeling. The model results matched well with the experiment and can be used for future local modeling of the trace geometry. The peak strains measured by the strain gauge in the cantilever at 40,000 g stress levels were found to be in excess of 10,000 με. FE analysis confirmed that the peak strain at the traces near the fixture was approximately 15,000 με.

Future work includes improving the AJP process to eliminate the bamboo cracking pattern as well as producing and testing more samples to provide Weibull statistics and parameters of the different printing methods under severe shock conditions and high strain rates. Furthermore, a local FE model is required to understand the effects of the local trace geometry on its reliability during the impact event and to develop a drop-durability model of the printed silver trace material using the stress, strain, and strain-rate values extracted from the local FE model.

• U.S. Army Research Laboratory (Award No. 160813; Funder ID: 10.13039/100006754).

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.