Caleb J. Reesea,
Grant M. Musgravea,
Jitkanya Wong
b,
Wenyang Panb,
John Uehlinb,
Mason Zadanb,
Omar M. Awartanib,
Thomas J. Wallin*c and
Chen Wang
*a
aDepartment of Materials Science and Engineering, University of Utah, Salt Lake City, Utah 84112, USA. E-mail: chen.wang@utah.edu
bReality Labs at Meta, Redmond, WA 98052, USA
cDepartment of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA. E-mail: tjwallin@mit.edu
First published on 12th July 2024
The continuous accumulation of electronic waste is reaching alarming levels, necessitating sustainable solutions to mitigate environmental concerns. Fabrication of commercial electronic substrates also requires wasteful high heat. To this end, we develop a series of reprocessible electronic substrates based on photopolymerizable polyimides containing degradable ester linkages. Five imide-containing diallyl monomers are synthesized from readily available feedstocks to produce high-quality substrates via rapid photopolymerization. Such materials possess exceptional thermal (thermal conductivity, K = 0.37–0.54 W m−1 K−1; degradation temperature, Td > 300 °C), dielectric (dielectric constant, Dk = 2.81–3.05; dielectric loss, Df < 0.024), and mechanical properties (Young's modulus, ∼50 MPa; ultimate elongation, dL/L0 > 5%) needed for flex electronic applications. We demonstrate mild depolymerization via transesterification reactions to recover and reuse the functional components.
Much research approaches the end-of-life issue in electronics by developing new polymeric electronic substrates (e-substrates) with inherent reprocessibility via chemical, thermal, or biological means. In this paradigm, complete degradation or removal of the e-substrate allows for the collection of expensive and reusable electronic components to minimize waste and amortize costs. However, these studies often simplify the circuits design. This simplicity often ignores the stringent material requirements for e-substrates in consumer devices (e.g. high dielectric constant and low dissipation factor for dense circuits and data transfer applications), realities of conventional large-scale device manufacturing (e.g. Pb-free solder reflow processing at ∼250 °C), and the economics of the hardware market (e.g. low raw material and processing costs). For example, several studies report electronic devices built on biodegradable substrates of modified fibrous11,12 or synthetic aliphatic polyesters such as polylactic acid13 and polycaprolactone.14–16 The degradability of these substrates typically comes from the ester linkages undergoing hydrolysis under suitable conditions (heat, oxygen exposure, microorganisms).17 Unfortunately, these polyester systems possess low heat stability well below conventional processing temperatures,18–21 and the conditions for hydrolysis can be so innocuous as to occur in ambient environments leading to premature device failure.11,22 Other work utilized liquid metal circuits printed on top of thermoplastic films to separately recapture the e-substrate, conductive traces, and components.23 Ignoring the inherent concerns with solvent and thermal resistance, thermoplastics are often susceptible to creep and plastic deformation which threaten dimensional stability. Additional shortcomings include that interchain ionic or hydrogen bonds can become active at high frequencies to the detriment of dielectric performance. Recently reported recyclable crosslinked polymers (or “network polymers”) may overcome the drawbacks with high thermal stability. However, the fabrication and reprocessing of these novel thermosets is often costly, requiring multi-step processing that involves solvent evaporation to form films.24–26 As an alternative, crosslinked polyesters can be formed efficiently at room temperature without solvents using mild reaction conditions.27–30
By comparison, polyimide (PI), particularly aromatic polyimides like Kapton®, is a popular substrate in electronic devices due to their unique combination of outstanding mechanical properties including tensile modulus ∼2 GPa; ultimate tensile strength ∼230 MPa; ultimate elongation ∼70%, thermal stability from −250 °C < T < 300 °C; relatively high thermal conductivity ∼0.12 W m−1 K−1; low coefficient of thermal expansion (CTE) ∼ 20 ppm °C−1; high glass transition temperature, Tg > 300 °C, high dielectric constant (Dk or εr), of 3.5 at 1 kHz; and low dielectric loss (Df or tanδ) of 0.002 at 1 kHz.31,32 Consequently, PIs are ideal substrates for robust, high density, miniaturized circuits, particularly since the superb electrical barrier performance persists in thin film form factors for flexible electronics. Unfortunately, the exceptionally stable imide bonds that help to impart these properties do not readily cleave under mild conditions, limiting reprocessing. Even with the addition of ester linkages, aromatic poly(ester imides) are similarly difficult to degrade.33–37 Another challenge associated with PI substrates is their synthesis. In a typical process, imidization of a polyamic acid solution occurs at exceptionally high temperatures (T = ∼250 °C) and once the resulting solid film forms, photo- and melt-processing becomes prohibitive. As a result, the conventional PI film processing complicates the build-up of multilayered circuit architectures (e.g. application of interlayer adhesives between PI films, laser ablation of PI to form vias, copper deposition for through-plane electrical conductivity) required for dense electronic devices.
Herein, we develop degradable polyimide photopolymers compatible with the existing materials, processing, and application requirements of commercial electronics. We propose using thiol–ene click chemistry as an ideal photopolymerization method for the design of ester-containing networks.38 Despite free-radical polymerization of acrylate or methacrylate free-radical polymerization being a common choice for photopolymerization, it suffers from oxygen inhibition, often resulting in tacky surfaces due to incompletely reacted monomers.39 On the contrary, thiol–ene polymerizations have the benefit of little oxygen inhibition and are proven capable of forming thin films with non-tacky, smooth surfaces.40 Further, many commercially available thiol monomers possess the degradable alkyl ester linkages, like trimethylolpropane tris 3-mercaptopropionate (TMPMP). With respect to thermomechanical properties, we envision imide containing alkene monomers boosting backbone rigidity, resulting in high Tg and high Young's moduli. In this work, we synthesize a family of ester-containing diallyl imides of varying structure derived from commercially abundant chemical feedstocks. We then demonstrate that the photopolymerized thiol–ene polyimide networks result in quality e-substrates with excellent dielectric properties. Depolymerization of these poly(imide ester) networks via transesterification reactions stimulated at ambient temperatures results in successful recovery and reuse of components.
As a preliminary attempt, we used pyromellitic dianhydride to synthesize a diallyl bisimide monomer, since it is one of the most common dianhydride feedstock for polyimide synthesis.43,44 However, the diallyl pyromellitic bisimide possessed a high melting point (Tm) of 220 °C, which recrystallized rapidly, preventing practical use as a liquid resin. Thus, we investigated other dianhydride feedstock chemicals, to obtain diallyl bisimide monomers with lower Tm's. As shown in Scheme S1,† we identified eight dianhydrides and synthesize their corresponding bisimide diallyl monomers, including 4,4′-oxydiphthalic anhydride, 4,4′-(4,4′-isopropylidenediphenoxy)bis(phthalic anhydride), cyclobutane-1,2,3,4-tetracarboxylic dianhydride, benzophenone-3,3′,4,4′-tetracarboxylic dianhydride, and 3,3′,4,4′-biphenyltetracarboxylic dianhydride, ethane-1,2-diyl bis(1,3-dioxo-1,3-dihydroisobenzofuran-5-carboxylate), 1,2,4,5-cyclohexanetetracarboxylic dianhydride, 4,4′-(hexafluoroisopropylidene)diphthalic anhydride. Dianhydride monomers similar to what are used in commercial PI films were explored in order to illicit similar high-performance properties that these films possess. These successful syntheses all resulted in yields over 70%. Three of the monomers were incapable of forming stable liquids similar to the pyromellitic bisimide. Therefore, we chose the remaining five, namely diallyl ester, diallyl ether, diallyl cyclohexane, diallyl diether, and diallyl hexafluoro (structures shown in Fig. 1 legend), which possessed Tm's lower than 160 °C (Fig. S1†). We focus the rest of the study on these remaining five monomers. We chose TMPMP as the thiol monomer to allow direct comparison of each diallyl bisimide monomer with sufficient crosslinking. In the following, we name the poly(imide ester) thiol–ene networks from the five diallyl bisimide monomers as PI-ester, PI-ether, PI-cyclohexane, PI-diether, and PI-hexafluoro.
In our formulation, we added the corresponding diallyl imide in 5 mol% excess to TMPMP (1.05 [CC]
:
1 [SH]), and 1 wt% diphenyl(2,4,6-trimethylbenzoyl)phosphine oxide (TPO) as the photoinitiator (Scheme 1A). The excess diallyl imide ensures a quantitative conversion of the thiol groups, since free thiols can react with metals.45,46 Minimal effects on physical properties were seen when comparing 5% allyl excess polymers to stoichiometric polymers formulations by Tg measurement (DMA) and tensile test. In practice, resulting liquid resins possessed a sufficient pot life (order of minutes or greater) such that transfer into heated (T ∼ 80 °C) silicone molds did not result in any noticeable phase separation or solidification. We then irradiated at 405 nm UV light (5 mW cm−2) and 80 °C for 10 minutes to polymerize the resin. Fourier-transform infrared (FTIR) spectroscopy confirmed complete disappearance of the thiol peak (2569 cm−1) and nearly complete disappearance of the alkene peak (3092 cm−1), shown in Fig. S2.† Monitoring the thiol peak by real-time FTIR confirmed the rapid polymerization kinetics typical in neat thiol–ene photopolymerization systems, achieving 63% thiol conversion within 10 seconds with 5 mW cm−2 light intensity (Fig. S3†). The resulting materials were found to be free-standing, glassy, tough, and transparent films, allowing us to readily handle and flex the films at ∼200 μm film thickness (Scheme 1B).
We quantified the thermal performance of these network polymers by TGA, DSC, CTE, and thermal conductivity. Similar to conventional polyimides, these polymers exhibit high thermal stability, with temperatures for 5 wt% degradation ranging from 368 °C to 385 °C (Fig. 1D and Table 1), far above conventional solder reflow temperatures (265 °C).47 The CTE of the glassy polymers is 74–79 μm m−1 °C−1. Typical polyimide flexible electronic substrates such as Kapton HN® possesses a CTE of 20 μm m−1 °C−1 at ambient conditions. CTE matching between substrate and mounted component is important to mitigate thermal stresses. It is important to note that all polymers exhibit an increase in CTE above their glass transition (see ESI†). Our PEIs also exhibit thermal conductivity ranging from 0.38 to 0.54 W m−1 K−1, which is higher than that of Kapton HN® (0.12 W m−1 K−1).48 Higher thermal conductivity is often desirable for heat dissipation. In fact, our neat photopolymers possess thermal conductivities similar to Kapton MT®, a commercial polyimide which contains additives specifically to improve its thermal conductivity (0.46 W m−1 K−1).49 Together, these thermal properties compare well with commercial polyimides and even outperform in the case of thermal conductivity.
One primary concern for dense electronic circuits is the dielectric behavior of the underlying substrate. Unfortunately, most of the extant literature on reprocessible polymer e-substrates do not quantify it adequately. We measured both the dielectric constant (Fig. 1E) and dielectric loss (Fig. 1F) of our polyimide network polymers at ultra-high frequency regimes relevant to wireless data transfer (Cellular networks, WiFi, Bluetooth, Ultra-wide Band, etc.) At 4.1 GHz, these photopolymers possessed dielectric constants (Dk) ranging from 2.69 (PI-ether) to 3.05 (PI-ester), slightly lower than that of Kapton-NH (3.4). We believe the relatively high dielectric strength of the PI-ester polymer originates from the polar ester groups and could be further accentuated by including high dielectric fillers. Dielectric loss (Df) values were found to range from 0.0145 to 0.0169, which is again comparable to commodity polyimides with dielectric loss values ranging from 0.0037 to 0.020.44 In summary, our materials possess similar dielectric behavior to Kapton, confirming our molecular design motif and further suggesting suitability in extant applications.
We monitored the depolymerization kinetics by weighing the residual mass of the polymer sample at varied durations. The PI-ester depolymerized the fastest, completing within 12 hours (Fig. 2A). Importantly, PI-ester samples showed little swelling, suggesting that the depolymerization reaction rate exceeded the rate of diffusion. Initially, the other four PI networks showed significant swelling—gaining mass during the first few hours as the solution infiltrates the network—prior to sufficient transesterification disrupting network percolation resulting in broken macroscopic films (Fig. S5†). We hypothesize that PI ester depolymerizes quickly due to the addition of two additional ester linkages within the imide monomer, which are absent in the other four monomers. Chemical analysis using 1H-NMR of the resulting small molecules confirmed the formation of methyl esters at 3.76 ppm shifting from the original mercaptopropionate peak at 4.03 ppm (Fig. S6†). The aromatic ester adjacent proton peaks also shifted from the ester monomer at 4.75 ppm to the degraded monomer's methyl ester location at 3.79 ppm. Thus, we conclude that transesterification with methanol is successful in depolymerizing these poly(imide ester) networks under ambient conditions into methyl ester containing molecules. Much work has been performed using transesterification to repolymerize degraded polyester polymers.51–56 We believe our degraded products are an ideal candidate for repolymerization in a similar manner. However, when similar procedures were used, side reactions occurred. Ongoing work to characterize and repolymerize the degraded molecules using transesterification will be reported elsewhere. As shown in Fig. 2A, the rate of swelling is readily adjustable by controlling the interaction between the solution and polymer (e.g. the cosolvent and its concentration).
![]() | ||
Fig. 2 Depolymerization of the PI-ester polymers as substrates, enabling recovery of electrical components. (A) Reaction kinetics show rapid loss of mass for the PI-ester at different concentrations of DCM and methanol using constant 0.07 M K2CO3. The PI ester can be seen to completely degrade within 12 hours with 50 vol% DCM. (B) Snapshots of demonstration of a functioning circuit comprising liquid metal (LM) wires, a battery, a resister, and an LED bulb used for a second life (i). The PI-ester polymer substrate was fully degraded in 2 hours (v), recovering the liquid metal wires and components. The recovered components were used in a new, functioning circuit (vi). It should be noted that the battery was removed before the polymer degradation. The video is included in ESI.† |
To demonstrate electronic component recapture via substrate depolymerization, we made films (100 μm thick) from the PI-ester based resin and added a simple routing circuit and basic components. For our conductor, we chose eutectic gallium indium alloys (EGaIn) liquid metal due to its mechanical compliance and more importantly, its recyclability.57 There is a variety of techniques to pattern liquid metal circuits, but in this study we choose selective wetting where a shadow masking combined with physical vapor deposition produces a 20 nm Chromium and 200 nm copper seed layer onto the film into the desired circuit pattern. Then the sample is dip coated into an EGaIn bath.58,59 EGaIn favorably alloys and wets this seed layer to form conductive traces that are deformable (see ESI† for more information). We completed the circuit by attaching electronic components: LED bulb, resistor, R2023 coin cell battery, and battery holder (Fig. 2B). After confirming the circuit is performant, we removed the battery and initiated recapture by immersing the device in the degradation (50% DCM, 50% methanol and 0.07 mol L−1 (M) K2CO3) solution at room temperature. After 2 hours, our PI-ester substrate depolymerized and we observed no visual change of the remaining electronic components. We collected the components, rinsed with isopropyl alcohol and let them dry. We repeated this process for three such circuits and noted that the resistors did not exhibit statistically significant changes in resistance: Ωavg, before = 5.17 ± 0.05 Ω; Ωavg, after = 5.23 ± 0.05 Ω (N = 3) (Fig. 2B-vii). To further confirm functionality after substrate removal, we placed these same components onto a fresh, patterned substrate. This “new” circuit is also performant, as demonstrated by the illuminated LED in Fig. 2. In conclusion, the poly(imide-ester) substrates allow for the fabrication of functional, flexible circuit and the recovery of the components by depolymerizing the substates.
Today, electronic devices using PI rely predominantly on pre-fabricated PI films due to the inability to liquid-process PI while staying within an acceptable thermal window of many electronic subcomponents. Linear, aromatic PI has prohibitively high Tm (>300 °C) that prevents melt processing, and alternatives based on liquid poly(amic acids) precursors still require imidization steps at >250 °C to initiate ring-closing reactions.43 In order to obtain dense electronic circuits with conventional (i.e., prefabricated PI films) approaches, one would need to build each electronic layer onto a film separately and then align and stack the films. Obtaining the necessary through-layer electronic connections is challenging, requiring additional subtractive steps to “drill” through the substrate to create a “via” and then applying suitable solder to achieve conduction. Electronic components placed on top of the films break 2D planarity and different components are often different heights. Applied liquid “build-up layers” can accommodate this height differential, but interlayer adhesion is problematic. Adhesives also create discrete transitions in stiffness, threatening overall flexibility and mechanical reliability.
By contrast, our resins are liquids at ambient temperatures, which permits straightforward deposition, casting, and leveling of thin layers iteratively. As shown in Scheme 1B, we can readily doctor blade and cure films down to 200 microns. Starting with a preformed PI film, we can place conductive traces and components conventionally. The thermal resistance of our PI films permits solder reflow of Pb-free solder at temperatures of 265 °C for 10 min (see Fig. 3A). For packaging, we can then cast more resin to planarize and insulate this electronic layer. In Fig. 3B, an ohmmeter probe on the bare conductive trace of the layer measured R = 0.8 Ω but after recoating and curing of our PI-ester the probe measures an open circuit (R > 108 Ω). Rather than insulating the entire circuit, we can also achieve “redistribution layers” for multilayered circuits. By employing photomasks, we can selectively photopolymerize our PI-ester in desired regions and leave unexposed regions uncured. This additive approach yields the necessary "via" directly; we can simply remove the uncured resin (University of Utah logo, Fig. 3A) and backfill the "via" with a conductor. The next layer's traces and components can then be built conventionally while simultaneously achieving the desired insulation and conductivity with the previous layer. This workflow is repeatable and infinite for an arbitrary number of layers of flexible circuitry. This proposed manufacturing approach significantly reduces the complexity of building multilayered PI based circuits.
Furthermore, we evaluated the dimensional stability of our network polymers. We expected the networks to resist creep deformation given their covalent crosslinking. Creep tests conducted on PI ester showed little strain using film samples in tension, performed at 25 °C, 100 °C, and 200 °C with 0.1 MPa of constant stress. It is worth noting that the dimensional stability at high temperatures is particularly of interest for solder reflow as mentioned above (Fig. 3A). At ambient, 25 °C, a low creep strain of 0.2% after 50 minutes was observed while at 200 °C a creep strain of 5% was found showing high temperature creep resistance. Much of this “creep” however was due to thermal expansion (3.1% from 25 °C to 200 °C) based on calculations using PI ester CTE.
Fig. 4A shows the supply chain of aromatic diamine, utilizing aromatic building blocks like benzene and phenol, with nitrogen introduction occurring via oxidation with nitric acid followed by reduction to obtain an aromatic amine (e.g., aniline).61 By comparison, Fig. 4B shows the supply chain of aliphatic amines, typically produced from ammonia exchange with halides, allowing a more straightforward supply chain due to the direct use of ammonia as the nitrogen source.62 Fig. 4B also illustrates the supply chain of multifunctional thiol monomers, where mercaptopropionates are derived from mercaptopropionic acid and a multifunctional alcohol by Fischer esterification, a process commercialized by several companies.63 The addition of hydrogen sulfide to acrylic acid produces mercaptopropionic acid, with increasing acrylic acid production from renewable resources making thiol monomers cost-competitive.64–66 Consequently, the poly(imide ester) photopolymer developed in this work is readily scalable, as it uses high-volume, low-cost building blocks.
Besides monomer costs, material costs include operational expenses. Fig. 4A shows that applying polyimide materials in device fabrication is energy-intensive, as aromatic polyimides have a high Tg (>300 °C) and require imidization between 200 °C and 300 °C over hours.31 In contrast, Fig. 4B depicts the workflow for photopolymerized substrates, which involves simple resin casting and photopolymerization in minutes. This process not only enables advanced manufacturing workflows via photopatterning but also significantly reduces time and energy input at an industrial scale.
It is important to note that the demands of some applications may require further improvements to the material performance and that while comparable, our materials fall short of Kapton along certain metrics. Future investigations could consider adding particulate fillers to further improve the thermal conductivity, thermal expansion coefficient, dielectric constant, etc. Such a straightforward strategy is likely to be successful as commercial polyimide films are often composites. Additionally, our materials are polyimides due to the incorporation of imide groups into our diallyl monomers. We could increase the polyimide character of the backbone by using oligomeric polyimides as precursors, though at the expense of hindered processibility such as increased melting temperature and viscosity. Additionally, we could manipulate the degradation rate by controlling the volumetric density of the ester groups decreases. Finally, future work could focus on further “closing the loop” by repolymerizing the substrate, though the economic motivation for such efforts remains unclear.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4lp00182f |
This journal is © The Royal Society of Chemistry 2024 |