Improved fatigue resistance in transfer-printed flexible circuits embedded in polymer substrates with low melting temperatures

Flexible electronics are of great interest and importance due to their applications in a range of fields, from wearable electronics to solar cells. Whilst resolutions of printed flexible electronics have been improving in recent years, there remain problems with mechanical fatigue and substrate cost, curtailing the use of such devices and resulting in increased cost and waste products. Here we present a novel method for improving the fatigue resistance of printed flexible electronics by a factor of ∼40 by sintering the electronics prior to transferring them into low-cost polymer substrates, such that they remain embedded. This method is demonstrated using circuits printed using silver nanoparticulate ink with an aerosol jet printer, and could be applicable to multiple different metallic inks. At the same time, this method can be used to transfer print circuits into polymers with low melting temperatures, without the need for otherwise detrimentally high sintering temperatures required for ink curing.


Introduction
Flexible electronics have gained significant interest in recent years [1,2], partly due to the rise in personal electronic devices such as folding phones [3], wearable electronics and skin inspired electronics [4][5][6]. One of the great advantages of flexible electronics is the potential scalability of manufacture. The world is currently facing a global silicon shortage [7], in part due to the slow batch processing for silicon electronics [8,9]. Owing to their flexible substrates, flexible electronics offer what is called 'roll-to-roll' manufacture, in a continuous rotating fabrication process [9][10][11][12]. This has significant advantages over batch production in terms of cost and scalability, which may help to solve significant shortages. However, flexible electronics currently struggle to achieve the resolution that has been developed in the silicon industry [4,9].
A major issue with most such printed electronics is that they are based on metallic conductive inks [27,28] which require post-deposition sintering. Sintering is a relatively high temperature process, which is often incompatible with the majority of polymers due to melting, softening or possible degradation of the polymers at typical sintering temperatures (∼200 • C). Attention is turning towards methods of preparing low sintering temperature inks to avoid these issues. These methods include the use of solution based inks, light based sintering methods and forming different complexes of metal nanoparticles [29,30]. However, sintering temperatures of 150 • C are still required even with such techniques. There are therefore relatively few polymers suitable as substrates for printed electronics, so commonly polyimide is used due to its thermal stability [6]. However, this has drawbacks as polyimide is a relatively expensive polymer (see supplementary table 1) which limits the potential applications of existing flexible electronics due to the prohibitive associated cost of the material. Furthermore, it is unsuitable for applications requiring transparent substrates. A viable method of using low cost polymer substrates for flexible electronics would open up novel new applications, such as environmental sensors including moisture sensors and even internet-of-things linked sensors [31].
Another major difficulty arises from the mechanical reliability of the flexible electronics, which remains a significant issue for the electronics industry due to the associated wastage and cost of replacement [32]. Part of the problem is that transferred or printed flexible electronic materials remain on the surface of the substrate, and are therefore more exposed and susceptible to damage. In addition, bending and flexing to relatively small of curvatures is a key feature of flexible electronics, which exposes the circuitry to fatigue damage [1,[33][34][35][36].
There is therefore a lot of interest in improving the mechanical behaviour of flexible circuitry [1,37]. One method to improve fatigue resistance is to reduce the thickness of the substrate, as this places the conductor closer to the neutral axis, where it experiences lower stresses and is therefore less vulnerable to fatigue [37,38]. Another possibility is to embed the conductive material inside the substrate, as this again moves it closer to the neutral axis, reducing the stress it experiences [4]. This has the added benefit of reducing the surface exposure of the conductor to wear and tear. For electronics that are embedded there remain some issues, chiefly that the encapsulation process can result in poor contact between the conducting material and substrate [17]. As a result, effective encapsulation of flexible electronics is gaining increased attention, primarily due to the environmental protection it offers [24,39], but also due to the range of functionalities that are made possible through the properties of the embedding polymer [40].
Some approaches to this have already been tried. Several successful attempts have been made to embed silver nanowires into different substrates via solutions of flexible polymers [41][42][43]. This approach is effective due to the high surface area to volume ratio of the silver nanowires and could be compatible with existing methods of printing for silver nanowires [44]. However, there are a couple of drawbacks: firstly, the weight loading of both the nanowires and the solutions are very low (typically less than 1%), resulting in low deposition rates and large wastages of solvent. Secondly the drop casting and spin coating used here are effective for the production of films but not for connections between different parts of a circuit, although it is feasible that printing methods may enable complete circuit production. Another approach by Frisbie et al have been based on reduction of copper based solutions using printed silver as a catalyst [45][46][47][48]. This approach does enable relatively robust connections to be formed, however it has limitations, as photolithographic techniques are used to form the initial master pattern for formation of the substrate, which increases the cost and difficulty of the process, as well as introducing a need for alignment of substrate and print.
There remain challenges to be solved in order to produce mechanically reliable flexible electronics, particularly at a low enough cost for disposable sensors. In this work, by separating the electronics sintering from the substrate, we present a novel method to embed aerosol-jet printed silver based electronics into low-cost polymer substrates. The resultant circuits are both flexible and wear resistant, with fatigue lifetimes more than 40 times longer than conventionally printed circuits. This method is shown to be compatible with a range of substrates and able to transfer complex designs into each polymer.

Methods
Aerosol jet printing (AJP) was used to produce the transferable circuits in this work with an Optomec AJ200 aerosol jet printer. Circuitry was printed via a silver nanoparticle (Ag NP) ink (Prelect TPS 50 G2, Clariant, diluted 1:1 with de-ionised water) as discussed previously [49]. The Ag NP ink was printed onto an aluminium foil substrate according to designs created in Autocad 2019. Following deposition the ink was cured on the aluminium foil by heating to 200 • C for 2 h. After curing, the foil was wrapped around a glass slide, with the circuit on the exposed surface as shown in figure 1, without the need for post-transfer high temperature sintering. The glass slide was placed onto a custom made stainless steel heating block with an internal k-type thermocouple that is accurate to ∼1 • C, alongside a clean glass slide. Pellets of selected polymers were placed on top of the circuitry and the temperature raised to approximately 30 • C above the melting point of the polymer, as described in supplementary table 1. The clean glass slide was positioned above the polymer melt and was approximately 100 kPa of pressure was applied by hand to create a film of polymer over the printed circuitry. The samples were then removed from the heating block and allowed to cool, which allows the polymer to encapsulate the printed circuit. Then the aluminium foil was peeled off the polymer, leaving behind an embedded circuit, as shown in figure 1. Suitable electrodes may then be attached to the surface using silver conductive paint to connect to an external controller. Free arc bending tests were carried out as described in figure 2(a) and Yi et al [1]. Each bending sample consisted of three separate parallel lines, each with contact pads to enable four-point resistance measurements. The same design was printed onto Kapton sheet and separately onto aluminium foil. The circuit was transferred from the foil into a melt of poly-l-lactic acid (PLA, molecular weight 85000-160000, Sigma Aldrich) as shown in figure 1(a). Both Kapton circuits and PLA circuits were attached to a linear motor that could oscillate between set distances of 42 mm and 28 mm, enabling the circuits to move between flexion and extension as shown in figures 2(a) and (b). The flexion induces a peak radius of curvature of ∼6 mm, which, according to a beam bending analysis using the Euler-Bernoulli equation [50], this would suggest a peak strain of ∼1.2% for the conventional samples and approximately 0.6% for the embedded samples. The resistance of each printed line was continuously monitored using a four-point measurement (Kiethley sourcemeter). A 2% increase in resistance was used as a failure criterion. This value is sufficiently high so as not to be affected by system noise, but also not so large that tests require an unreasonable time to complete. A total of 5 samples of each type were tested, each with 3 printed traces, giving 15 measurements for both transferred and conventional methods. This enables a reliable method of measuring the durability of the circuit, as the resistance of the circuit was found to increase over multiple cycles (figures 2(d) and (e)).
Rubbing tests were carried out using a custom linear motor set up. A custom circuit (shown in figure 3) was printed using the aerosol jet printer onto both polyimide sheet (125 µm, Goodfellow) and aluminium foil. The circuit on foil was transferred into a melt of polyvinylidene fluoride as described above. The circuit consisted of eight parallel traces, connected at one end to a common voltage source and to separate Arduino input pins at the other. The sample was fixed to a glass slide with adhesive and then positioned in the linear motor set up. A hardened chrome steel ball (5 mm diameter) was positioned above the sample, and repeatedly dragged across the eight printed traces at 2 Hz with a normal force of 2 N (assuming a coefficient of friction, µ of ∼0.2, this corresponds to a frictional force of about 0.4 N). Contact pads either side of the test traces detect each pass of the ball bearing, allowing a the Arduino to count the number of cycles (as shown in figure 3(a)). After each cycle, the input pins connected to each test trace are poled, detecting if the circuit is still intact or if the conducting path has been broken. The lifetime of each printed trace is taken as the number of cycles before the open circuit condition is reached. A total of 4 samples of each type were tested, each with 8 printed traces, giving a total of 32 lifetime measurements. Scanning electron microscopy images were taken using a Hitachi TM3030Plus.

Results
AJP printed silver ink on aluminium foil was found to be suitable for transfer, via the method described in section 2, to a range of polymer substrates, as described in supplementary table 1. The transfer method presented in this work is therefore compatible with a range of polymers which would not survive the high temperatures required to sinter the Ag NP ink without this method. When transferred into a polymer, the printed silver is encapsulated, with only the bottom surface of the print exposed, as shown in figure 1(b). The encapsulation process largely protects the printed silver from exposure to wear, as only the surface remains exposed to damage. Furthermore, the encapsulation of the silver in a thin layer of polymer may reduce the thickness of a device, and position the silver closer to the neutral axis of bending, where it is exposed to smaller strains than on the surface, as is the case in conventional printed electronics.
The effect of the printed trace height (thickness) on effective transfer is shown in figure 1(c). The maximum line thickness successfully transferred was ∼100 µm, as greater thicknesses collapsed under pressure of the gas flow during printing. However greater thicknesses may be achievable with a smaller aspect ratios of prints (increased width) or different inks. Supplementary figures 4 and 5 also demonstrate that the overspray associated with aerosol jet printing is also susceptible to transferring into the polymer substrate, which may limit the achievable resolution.
In order to test the flexibility of the transferred circuits, free arc bending tests were carried out as shown in figure 2(a). Circuits printed directly onto polyimide and those transferred from aluminium foil into PLA were compared. The resistance of the printed lines was measured during oscillation between extension and compression as shown in figure 2(d). Bending tests were found to increase resistance under compression (with the electrodes under strain) and decrease again during extension, with the electrodes once again relaxed due to the formation of microcracks within the printed silver lines (supplementary figure 2). Over the course of hundreds to thousands of cycles the resistance increased across the cycle, as shown in figure 2(e). This enables resistance to be used as a measure of the lifetime of the samples. The failure point may be defined in multiple different ways [51,52], however, it is common for the failure threshold to be defined at some value of normalised resistance [37,[51][52][53]. In this work we have defined the failure threshold to be when the resistance of a sample, R, reaches 1.02 times the initial resistance, R 0 , thus allowing comparison between samples with varying values of R 0 . This threshold was chosen as it represents a reasonable increase in resistance that can be achieved within the timeframe of the experiments, and that still differentiates well between samples. The distribution of sample failures for bending tests conducted on conventionally printed samples and embedded printed samples is shown in supplementary figure 1.
It was found that the embedded silver had ∼50 times the mean lifespan of the conventional printed silver (5340 cycles compared to 98), as can be seen in figure 2(c). The means and medians of the populations were tested and found to be significantly different at the 1% level (two sample t test, p = 3.57 × 10 −13 , Mood's median test, p = 8.39 × 10 −7 ). Supplementary figure 2 demonstrates the failure mechanism of these samples, with hairline cracks forming perpendicular to the bending direction. The hairline cracks were the main visible evidence of fatigue and are likely to be responsible for the increase in resistance associated with sample failure, although in the embedded samples, failure of the substrate itself is likely to be a contributing factor, as in supplementary figure 2(c); this may be due to the decreased flexibility of the PLA compared to the Kapton sheet.
Following bending tests, rubbing tests were also carried out to examine the wear resistance of the embedded electrodes in response to surface damage. These tests were conducted in the manner shown in figure 3(a), and as described in the methods, with a steel ball being used to provide the wear. The results of the tests are shown in figures 3(c) and (d). In the conventionally printed sample, as a result of the printed silver rising above the surface of the substrate (figures 3(d)(i)), rubbing eventually causes debonding at the interface between the silver and substrate, which results in complete removal of the silver in the vicinity of the contact area (figures 3(d)(ii)). Whereas, in the embedded sample, the adhesion between the silver and polymer is sufficiently strong to prevent debonding. The distribution of sample failures for rubbing tests conducted on conventionally printed samples and embedded printed samples is shown in supplementary figure 3. The embedded samples had ∼40 times greater mean lifespan than the conventionally printed samples (1820 cycles compared to 45). Furthermore, both the mean and median of the populations were found to be significantly different at the 1% level (two sample t test, p = 2.67 × 10 −27 , Mood's median test, p = 5.24 × 10 −14 ). For the embedded samples the silver can still be damaged, yet it is not vulnerable to complete removal (figures 3(d)(iv) and supplementary figure 4). From figures 3(d)(iv) and supplementary figure 4 it is also possible to observe that in the embedded sample, wear of the surface leads to redistribution of silver into the surrounding area.
In addition, the lifetime of the prints was found to increase with increasing layer thickness, as shown in supplementary figure 5. This increase plateaued at greater thicknesses of print, this is likely due to the aspect ratio becoming significant and therefore the wear tending to a limiting value. The resistance of the printed lines was found to decrease inversely with the number of print layers (equivalent to area) and increase linearly with length according to Pouillet's law [54], as shown in supplementary figure 6.

Discussion
The applications of the transfer technique presented above are wide ranging. It is perfectly feasible to replace the majority of flexible electronics manufacture, which are usually based on printing [9], with this low cost technique, with relatively small changes in manufacturing processes. This would help to prevent debonding, which remains a major issue for flexible electronics [36], as well as improve the fatigue resistance of manufactured devices, which is currently lacking. One of the major advantages of the method presented here is the separation of sintering from the end substrate. By curing the silver ink on the aluminium foil prior to transfer, low melting point, low cost polymers, for instance polyethylene, may be used as the substrate instead of costly high temperature polymers such as polyimide. Furthermore, this will also reduce the requirement for inks designed for low temperature sintering, which are often expensive. As a result, this process could also enable a variety of new devices to be implemented which would not have been possible before due to prohibitive cost [5]. Novel sensors may also be designed to utilise the properties of the given polymer base. For example, a soluble polymer base may be used to create low cost, flexible moisture sensors. As the polymer dissolves, the circuit will be released and quickly break down, turning off the current through the circuit. This could have applications in moisture sensitive areas, such as in food and chemical storage [31], similar devices could also be produced with sensitivity to certain solvents, or others may utilise UV sensitivity to determine sunlight exposure. Furthermore, transparent polymer substrates could be used in applications where this may be a requirement.
One question that remains, is why the novel transfer method presented here is successful with a variety of different polymers, but not all. Certain polymers are able to encapsulate silver effectively and de-bond it from the aluminium foil substrate while others remain attached. There are several important factors governing the success of a given transfer: (i) the surface energies of the polymer-silver interface and silver-aluminium interface; (ii) relative thermal expansion coefficients of silver and polymer; (iii) the melting point of the polymer; and (iv) the viscosity of the polymer during melting.
The relative surface energies of the polymer and aluminium interfaces are likely to have an effect as they will determine the adhesion of the different components. The combination of melting point and thermal expansion coefficient is also likely to play a role, as a greater contraction when cooling from the melt will cause residual stresses, which may provide a stronger frictional force between the silver and polymer. Finally the melt viscosity is also likely to be important, as this will affect the contact area of the polymer-silver interface, with a low viscosity melt forming a greater contact with the surface. Supplementary table 1 shows values for the surface energies, thermal expansion coefficient and melting point of various polymers. There appears to be little in the way of a clear link between these parameters and the success of transfer, and it is likely that the true determination of success will be as a result of a combination of the above factors that is beyond the scope of the work here.

Conclusion
In this paper we have produced an effective, low cost, low temperature method for improving the fatigue lifespan of printed flexible electronics. The method has been shown to increase lifespan against multiple different sources of damage, by a factor of at least 40, with some evidence to suggest that the limiting factor for our embedded electronics is now the substrate polymer. Importantly, this technique allows inks requiring high sintering temperatures to be deposited within polymer substrates with low melting temperatures, potentially both reducing the cost of flexible electronics, as well as enabling novel uses. This technique is compatible with a range of different polymers, and it is expected that there are many more for which it could be used, enabling a variety of new uses for flexible electronics. Such uses may include new sensors that exploit the polymer's properties to detect a given chemical or environmental change. For instance, a water soluble polymer with transferred electrodes could act as an effective moisture sensor, when such a polymer is exposed to water and dissolves away, the electrodes would be exposed and the circuit is liable to be broken. Other interesting future work could include the application of this method to complex prints, including those in three dimensions, the use of it in novel sensors, such as those described above, and an effort to understand the factors influencing the success of the transfer technique, as there is currently no clear method of predicting the success of transfer into a given polymer. It is envisaged that, in the future, the versatility of the transfer printing process shown here may solve robustness and reliability issues for all manner of printed flexible electronics.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https:// doi.org/10.17863/CAM.96599.