Behavior of Polymeric Substrates and Sintered Silver Printed Hybrid Electronic Assemblies Subject to Extreme Acceleration Levels and Elevated Temperatures

This section provides the background and motivation for the present work and presents a review of the relevant literature.

Ever since the development and use of primitive printed circuit board (PCB) electronics in the 1950s and continuing into the complex heterogeneous integration substrates for integrated circuit (IC) packages of today, industries such as automotive and aerospace have sought to employ electronics in demanding mechanical shock environments where acceleration levels can exceed 80,000–100,000 g with pulse durations as long as 1–3 ms and velocity changes well over 300 m/s. For reference, test standards for typical consumer products are designed to accommodate accelerations at 2900 g with similar pulse durations but velocity changes no higher than 5–6 m/s [1].

Decades of research have shown that the challenges inherent in designing electronics for these environments can be overcome, typically through the use of stiff structures and circuit boards, careful component selection, underfill use, and potting of individual boards and complete packages. Indeed, conventional electronic packages appropriately designed and fabricated are consistently reliable and survivable in these extreme environments for various applications [2,3].

Changing operational applications, however, predict that electronic packages of the future will be required to withstand not just mechanical shock, but to do so at elevated temperatures. While the body of research concerning performance of electronic packages at elevated temperatures is robust [4], the combination of mechanical shock with elevated temperature offers significant challenges to the electronic designer since the very strategies that promote reliability for shock often degrade reliability at elevated temperatures. Nevertheless, limited efforts are underway to consider how conventional electronic systems respond to these combined loading scenarios [5].

In recent years, the increasing development of printed hybrid electronic (PHE) assemblies has led to growing interest in employing PHEs in these aforementioned extreme environmental conditions [6]. PHEs combine elements of both advanced three-dimensional-printed technologies for substrates and circuits with CMOS-based ICs. This combination allows the designer to leverage the advantages of both technologies, offering advantages over conventional PCB-based electronic packages in geometry, size, weight, capability, flexibility, and many other metrics [7,8]. This work will specifically focus on the reliability of embedded traces in PHEs printed on curvilinear polymeric substrates when subject to mechanical shock and elevated temperature.

While the potential advantages of PHEs are promising and alluring, there is a great diversity of PHE technologies [9,10]. This study focused on the use of polymeric substrates (not conventional FR4) and extrusion-based silver paste printing.

The most common polymeric substrates for use in PHEs are polyethylene derivatives, typically for applications requiring flexible substrates. All of these polymers except polyimide, however, have glass transition temperatures below 143 °C. Polysulfone is one of the highest-temperature polymeric substrates with a glass transition temperature of ∼188 °C, motivating its use for this study.

The preponderance of printed trace research is based on using aerosol jet technologies [8], but traces printed with aerosol jets struggle to perform well in extreme acceleration environments [11]. Direct ink writing/extrusion printing mitigates some of the concerns of aerosol jet printing, mostly because the output trace is thicker. Just as with any printing technology, process parameters have considerable influence on output quality and performance, and ongoing work is looking into these relationships [12]. Other related research is looking into ink-jet technologies for silver printing [13], the use of printed silver for interconnects [14], and the mechanical strength of sintered silver joints as a function of interconnect area [15].

Embedded components are of interest in this study because the use of polymeric substrates simplifies the design and manufacturing process thereof. While the use of embedded passive components is documented in literature, this work is done with conventional FR4 substrates [16]. Understanding the performance of polymeric substrates with embedded traces of printed and sintered silver subject to mechanical shocks (with or without elevated temperatures) is a gap in current research.

The focus of this section is the use of drop towers or drop testing to test the robustness of electrical traces under high strains and high strain rates. Most of the relevant past research specifically focused on electronics reliability falls into one of two categories—considering either failure modes within individual components that are mounted to traditional PCBs, or looking at complete IC packages at the board level. Considerable efforts have also been made in studying the effects of mechanical shock on complete electromechanical products.

Use of the JEDEC standard for mechanical shock testing [1] is the basis for most board-level reliability research, such as those concerning crack propagation in interconnects [17] and the characterization of solder interconnect materials' response at high strain rates [18]. Modeling and simulation examples are common [19–21], and also include strategies for predicting survivability [22]. Most of this work, however, is conducted at acceleration levels at or well below 10,000 g. Relatively few publicly available papers have been published for mechanical shocks at acceleration levels at and above 50,000 g.

The behavior of commonly used lead-free solders such as SAC 305 at elevated temperatures and subject to mechanical shock has been studied extensively by Lall et al. [10,23], Mattila et al. [24], and Yuan et al. [12]. Additionally, the shock performance of low-temperature tin-bismuth solders has been studied [25]. Several studies have assessed the reliability of printed silver at elevated temperatures, but this is generally in the context of isolated interconnects, not continuous traces [16], or at temperatures that are outside the scope of this effort [26]. None of the studies here described are for PHEs, a significant research gap.

Details of the test specimens, test setup, and test method are provided in this section.

The bidirectionally curved substrates of interest in this study were dome specimens that measured 40 mm in diameter and 36 mm in height. The domes had threaded bases for installation into fixtures, short vertical walls, and sharpness factors of approximately 0.5 (nearly perfectly hemispherical). The selection of domes (versus planar geometries) was driven almost exclusively by the desire to print electronics on curved surfaces. Additionally, this study provided insights into the additional rigidity of dome specimens under shock loading, compared to flat geometries. Future work will focus on the drop response of flat specimens.

The dome substrates were manufactured using injection molding. The primary dome material used for the high-temperature tasks in this study was polysulfone, selected based on its toughness and stability at high temperatures. Polycarbonate substrates were primarily used for the trace characterization portion, however. Material properties for these substrates, needed for finite element modeling, were based on consolidated information from previously published works and research efforts. These substrates formed the basis for electronic circuits of embedded traces and external connections. Examples of the blank and populated domes used are shown in Fig. 1.

The process for modifying the standard domes into populated test specimens is included in the Appendix. A depiction of the steps for this process is shown in Fig. 2.

The test matrices in Table 1 list the specific samples used with nominal values of acceleration and temperature (tolerance limits for the acceleration and temperature levels were +5000 g/−10,000 g and ±10 °C, respectively). Temperature values were selected based on physical capabilities of the substrate materials; acceleration levels for repeated drops were selected to balance data collection on specimen performance and test duration. The numbers in these tables refer to number of test specimens, where each specimen contained two individual circuit elements and so the actual number of circuits assessed is double the specimen numbers. The total number of specimens used were six blank domes and 20 domes with two traces each. Additional domes were used for destructive testing (cross-sectioning and physical analysis). In these tables, test points marked with a “*” had at least one specimen instrumented with an attached strain gauge. Strain gauges were adhesively mounted to the interior face of the dome, located at the bottom, with the axis of the strain gauge oriented vertically (towards the top pole of the dome).

Incoming test specimens were inspected, cataloged, and instrumented for strain and resistance measurement and electrically characterized for initial performance baseline. A few selected specimens were further characterized using optical microscopy, scanning electron microscopy, and computerized tomography scanning, to better understand the fabrication quality.

Once all pretesting steps were complete, the specimens were tested in accordance with the test matrix. Specimens were installed into fixtures and mounted to the drop tower. The drop tower used was a Lansmont HSX23 accelerated-fall shock test machine equipped with a dual mass shock amplifier (DMSA). The tower, associated equipment, and DMSA are shown in Fig. 3. At peak height, this drop tower provided a half-sine acceleration pulse with maximum acceleration of 100,000 g, pulse duration of 0.04 ms, and maximum velocity change of 23 m/s. The minimum acceleration level used was 25,000 g with pulse durations of ∼0.10 ms and velocity change of 12 m/s. Pulse duration decreases and velocity change increases with increasing acceleration. These relationships are well understood as shown in Figs. 4 and 5. While the specific values were unique to this drop tower and pulse programing, understanding the relationship trends between peak acceleration, pulse duration, and impact velocity are critical for mechanical shock testing engineers.

Specimens were mounted to the drop tower (specifically the DMSA table) by means of custom fixtures. For domes, the bottom threaded directly into matching threaded holes in the fixtures. For room temperature tests, the fixture was designed to simultaneously test a maximum of four domes (practical considerations generally limited this to three at once, see Fig. 6). At elevated temperatures, only one dome was tested at a time.

Table acceleration was measured by means of one of more accelerometers mounted to the table directly or the intermediate fixture. Fixtures were sufficiently rigid to ensure measurement accuracy regardless of accelerometer position. Accelerometers used included both piezoelectric (PCB 350B01 100,000 g, 350C02 50,000 g) and piezoresistive (Endevco 7270A 200,000 g) technologies. Accelerometers were mounted in accordance with manufacturer recommendations. The acceleration pulse peak and overall shape was consistent across all drops. Only the peak acceleration value was used for subsequent data analysis. When appropriate, drop tests were also instrumented with high-speed cameras. The standard high-speed camera was a Photron Fastcam SA-Z, and the typical frame rate used was 20,000 frames per second.

Elevated temperature testing was accomplished by means of a “furnace” installed onto the drop tower. This allowed for in situ heating of specimens after they were installed into drop tower fixturing. Heated specimens were dropped directly from within the furnace. Heating the specimens while they were already on the tower (as opposed to separately in a bench-top furnace) had two benefits: first, controlling the temperature of the specimen at the moment of drop was easier, and second, it avoided the need to move a hot specimen into position and install it within the tight confines of the DMSA assembly. A hot air gun was used for heating.

Heating levels were controlled manually, using a FLIR SC6100 thermal camera for wholistic temperature measurement of the specimen(s) before and after impact. Future work will use automated temperature control capability. Additionally, thermocouples were used for local temperature measurement, both to double-check thermal camera measurements and to guard against overheating of the fixturing or DMSA. Thermocouples were also used for camera calibration. The thermal camera had the resolution to show the thermocouples and take temperature measurements thereof. Emissivity values were not calculated; therefore, no other thermal image correction was performed. The thermocouples were mounted directly to the top surface of the specimens during testing.

Test results for substrate and sintered silver trace performance are provided in this section.

Characterization of the substrate was performed for two reasons: (1) to assess mechanical shock survivability and (2) to obtain material property data for use in finite element modeling efforts. The experimental characterization and assessment of the substrate material was facilitated since the blank substrates were relatively cheap and easy to manufacture. Drop survivability assessment was conducted by dropping blank specimens in the tower at increasing acceleration levels until failure. Polysulfone domes cracked after 3–4 drops at 100,000 g, and an example of a cracked dome is shown in Fig. 7. At 75,000 g, the polysulfone domes survived 40–50 drops before failing in the same way.

Strain gauges were mounted to several of the blank specimens to record strain as a means to both (1) assess peak strain and compare against previously published material property data and (2) generate baseline strain profiles for subsequent use in modeling.

Figures 8 and 9 show the first 1 ms of the strain profile at 25,000 g, 50,000 g, 75,000 g, and 100,000 g for polycarbonate and polysulfone domes, respectively. Maximum instantaneous strain rates for these different tests ranged from ∼420 to 860/s. Peak strain value and maximum instantaneous strain rate both increase with increasing acceleration level. For the 50,000 g and 100,000 g acceleration level, the Fourier transform of the strain response is shown in Fig. 10.

Elevated temperature testing of blank specimens was accomplished by heating specimens inside a specially designed furnace mounted to the drop tower. Through careful selection of the variables such as heater fan speed, heat setting, nozzle orientation, and positioning, the desired temperatures were achievable and maintainable. The system did, however, require careful monitoring to ensure even specimen heating.

Due to spatial temperature gradients, both maximum specimen temperature and mean specimen temperature were recorded at the moment before drop. Both substrate materials were tested at nominal target temperatures of both 87.5 °C and 150 °C.

As expected, integrity of the polycarbonate specimens was acceptable at 87.5 °C testing but compromised at 150 °C (the glass transition temperature of polycarbonate is ∼147 °C). Since the glass transition temperature of polysulfone is ∼188 °C, minimal damage was observed for these specimens. Figure 11 shows thermal camera images from nominal 150 °C testing.

The domes were instrumented with strain gauges. Strain gauge data for polycarbonate (solid lines) and polysulfone (dashed lines) domes as a function of temperature for a constant 25,000 g acceleration level are shown in Fig. 12. While the behavior of the material was consistent for all temperatures, the peak strain values for polycarbonate at 25 °C, 87.5 °C, and 150 °C were 2552 μm/m, 2740 μm/m, and 2911 μm/m, respectively; with the elevated temperature strains increasing 7.3% and 14.1% over room temperature strain, respectively. Similar trends were observed for polysulfone domes albeit with higher peak strain values; strain increases of 19.1% and 33.7% over room temperature levels were observed at 87.5 °C and 150 °C, respectively. We expected to see this type of strain increase given the softening of the polymeric substrate at elevated temperatures.

The standardized dome specimen for trace performance testing was shown earlier in Fig. 1. Each dome contained two independent traces. As described earlier, resistance across the trace(s) is measured by attaching lead wires to the pads. Multiple specimens were tested with multiple drops at fixed acceleration levels. To compare performance of the different specimens across the various acceleration levels, “input drop energy” was used as a metric for performance assessment. Drop energy was calculated for each drop by first determining velocity change. This was estimated by integration of the acceleration profile between bounds determined by 10% of the peak acceleration value in accordance with industry standard. This calculated velocity change compared favorably with observed velocity changes from high-speed video. A normalized (per unit mass) relative input drop energy was then estimated by squaring this velocity change, and cumulative input energy was determined by summing the impact energy for all drops of a given specimen.

Long-term electromechanical performance of the trace was measured by dropping specimens multiple times at fixed acceleration levels and measuring the resistance across the trace over time. The method of resistance measurement was kept consistent for all measurements. Failure was assessed to have occurred when the steady-state trace resistance doubled from the as-received pre-impact state. Figure 13 shows the result of these tests. The traces proved to be resilient at all acceleration levels. All traces performed similarly, with a slow slight degradation followed by a sudden (but not instantaneous) deterioration to failure. Failure occurred at 1 × 10⁶–3 × 10⁶ J/kg cumulative input energy for the 25,000 g tests and 50,000 g tests. The behavior of samples at 75,000 g and 100,000 g matched the curve shape, but the failure energy decreased to 5 × 10⁵ and 1 × 10⁴ J/kg, respectively. These cumulative input energy failure levels correspond to ∼150–170 drops at 25,000 g, ∼100–120 drops at 50,000 g, ∼50–70 drops at 75,000 g, and ∼35–45 drops at 100,000 g.

While the results of long-term degradation testing are interesting, some short-term changes are also worth noting. Because the acceleration pulse was so short and the strain rates so high, we observed instantaneous, yet significant, changes in resistance across the trace during the shock event. These transient changes were not observable when measuring resistance of the circuit at rest. Figure 14 shows how resistance (blue dots) is unpredictable during the impact event, but the high resistance values (25× the steady-state values) indicate that circuit performance was compromised during the acceleration pulse.

The precise cause of this circuit degradation was not definitely proven. The progression of cracks within sintered metals seemed the most probable cause, and cracking was observed in some specimens. Examples of this are shown in Fig. 15. These cracks were not universally observed in all traces, however. Some were observed to occur at areas of stress concentration such as the intersections of perpendicular traces.

Both polycarbonate and polysulfone substrate materials had favorable performance results when subject to mechanical shock. At room temperature, the polycarbonate domes never failed with 10+ drops at 100,000 g and hundreds of drops at lower acceleration levels (peak strain ∼10,000 microstrain). The polysulfone domes failed after 3–4 drops at 100,000 g and never failed after tens of drops at lower acceleration levels. Polysulfone substrates retained their survivability performance at elevated (87.5 °C and 150 °C) temperatures. Absolute strain values and strain rates were comparable at all temperatures. Polycarbonate substrates, however, softened with increasing temperature and were not a good material choice for elevated temperature testing, with absolute strain values increasing 7.3% and 14.1% over room temperature strain levels at 87.5 °C and 150 °C, respectively. Geometrical tolerances also could not be assumed constant once polycarbonates were exposed to elevated temperatures.

Sintered silver traces were highly survivable mechanically, never detaching from the substrates or otherwise experiencing catastrophic failures for any substrate and for any acceleration level. As expected, silver traces were subject to incremental damage caused by repeated exposure to mechanical shock. During a shock event, instantaneous intermittent continuity failures were observed. These were caused by cracking in the trace “opening up” under strain and lasted for about 0.1 ms. Additionally, traces degraded progressively and incrementally over time with the resistance increasing slowly until about 10⁶ cumulative impact energy. Both of these behaviors worsened with higher acceleration levels and more drops.

The testing methodology was robust, and desired acceleration levels were easily, predictably, and repeatably achievable. High-temperature drop testing was possible with the combination drop tower/furnace assembly. This was not a challenging modification to make and did not affect standard drop tower operation in any capacity.

In general, the use of embedded traces shows promise for many different applications. Sintered silver is tolerant to mechanical shock, especially when encapsulated in a dielectric layer. Both polycarbonate and polysulfone are effective substrates for embedded components. Polycarbonate is recommended for the most extreme duty applications requiring high strength, whereas polysulfone is the recommended material for any high-temperature (>100 °C) applications.

Further efforts will focus on (1) robust finite element model development, facilitating performance assessment and failure prediction for a wider range of substrate materials and geometries and (2) assessment of embedded components and component interconnects.

• DEVCOM Army Research Laboratory through the Center for Advanced Life Cycle Engineering at the University of Maryland, College Park (No. W911NF-23-9-0005; Funder ID: 10.13039/100019923).

The authors declare that they have no conflict of interest.

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

The process for modifying the standard domes into populated test specimens is included here:

Substrates: The substrates were modified in several ways from the as-molded configuration. Aside from the drilling of holes and other features required for external connections, the primary change was the use of a five-axis computer numerical control machine to perform “mill-and-fill,” a process by which small channels were milled into the surface of the dome. These channels were then used as guides for the printing of traces and the emplacement of components. This mill-and-fill process was important because it provided a known surface for printed traces (as the extrusion process is highly dependent on surface geometry and tolerancing), helped protect the traces during handling, and (in the future) will allow for components to be embedded within the substrate rather than mounted on the surface.

External Connections: External connections were required in these test specimens to monitor circuit resistance in real-time. All domes used the standard “H-pattern” two-trace layout, with four pads per trace for four-probe resistance measurement. First, 1.0 mm diameter holes were drilled through the dome substrate. Two holes were drilled for each of the eight trace connection pads, one on each side. Through these holes were fed ∼15 cm lengths of thin (0.25 mm) diameter single-strand copper wire. These eight sets of wire were pulled tight to secure, then secured to the dome surface using thin strips of masking tape. The monitoring wires were threaded through the holes and secured to the pads using either silver epoxy or low-temperature printable tin-bismuth (Sn42Bi58) solder, either extrusion-printed by machine (for small components) or by hand (for larger components or lead wires). The bonding was completed at 150 °C (to cure the epoxy or reflow the solder). Additional lengths of thicker copper wire were used to connect to the data acquisition system.

Traces: The channels that contain traces were 250 μm wide and each layer of trace material was 50 μm deep. Traces were extrusion-printed silver paste. After deposition, the traces were thermally cured per manufacturer recommendation (typically 135 °C for 20 min). In cross section, they were 250 μm wide (constrained by trench width) by 50 μm tall (with machined top). The dielectric was liquid photopolymer with clear UV cure. This dielectric had good hardness and good adhesion. Selection of the specific dimensions for trace and dielectric geometries was based on both functional (compatibility with circuitry, conductivity, etc.) and fabrication process parameter optimization (milling, printing, sintering, etc.) requirements.