Review: Solution processable liquid metal nanodroplets by surface-initiated atom transfer radical polymerization

David Schlachter, 2019 [Cite this article] (BibTeX)

Introduction

This review will focus on a 2019 article published in Nature Nanotechnology by Yan et al. which describes the compatibilization of a metal-polymer system (Yan et al. 2019). The metal is eutectic Gallium-Indium (EGaIn) and is a liquid at ambient conditions (melting point ≈ 16 °C). Previous work has explored applications of similar blends formed by mechanical mixing. To compatibilize their polymer-metal blend, Yan et al. avail themselves of a recently-developed tetherable polymer initiator which uses atom transfer radical polymerization (ATRP).

In this review, the background and context of the article by Yan et al. will be explored to demonstrate the challenges and advances contributing to their 2019 paper. First, the special properties of EGaIn will be explored, followed by a discussion of its application in self-healing circuits. ATRP will then be briefly reviewed, along with the development of the tetherable initiator. Previous attempts at compatibilization of EGaIn-polymers blends will be considered. Finally, the 2019 paper by Yan et al. will be reviewed in detail, and opportunities for future work identified.

Liquid metal alloys

Since discovery of indium and gallium in 1863 and 1875, respectively (Holden 2019), it has been possible to create alloys of these metals with melting temperatures near ambient conditions. The first investigation of low melting temperature gallium-indium alloys appears to have been conducted by the discoverer of gallium, Boisbaudran, in 1885 (Boisbaudran 1885). In a brief communication, Boisbaudran describes the formulation of four alloys having ratios (molar basis) of (1) 2In + Ga, (2) In + Ga, (3) In + 2Ga, and (4) In + 4Ga. Exact determination of the melting temperatures of the alloys was difficult due to an observed coexistence of crystal and liquid domains over a range of about 40 °C, starting at the melting temperature. All alloys were solid below 16 °C, began to melt around this temperature, and were fully liquid near 60 – 80 °C.

The melting points of pure gallium and of pure indium are 29.8 and 156.6 °C, respectively (Haynes 2011). Binary alloys of gallium and indium manifest a thermodynamic phenomenon referred to as a eutectic point, meaning that the melting temperatures of alloys decrease over the range of compositions relative to that of the pure components. A minimum melting temperature is obtained at the eutectic composition and is lower than that of either pure component. Of note, a eutectic alloy has a single melting temperature (rather than a range), and the solidified metal has one phase (the metals are ‘miscible’). The phenomenon of eutexia was described and named by Guthrie in an 1884 communication (Guthrie 1884). Guthrie’s work at the time was principally in the study of hydrated salt solutions, thus forming a major part of the article. However, the discussion of eutexia in salts is prefaced by a description of known eutectic metal alloys, which included bismuth-lead-cadmium, bismuth-tin, bismuth-lead, and bismuth-zinc. As an example, the melting points of pure bismuth and of pure lead are, respectively, 271.4 and 327.5 °C. The eutectic bismuth-lead alloy described by Guthrie had a composition (mass basis) of 55.6% bismuth and 44.4% lead, having a melting temperature of 122.7 °C.

For gallium-indium, the metal alloy used in the 2019 paper by Yan et al. (Yan et al. 2019), the eutectic composition is about (atom basis) 14% indium and 86% gallium, with a melting temperature of about 16 °C. These figures are taken from a 1991 review (Anderson and Ansara 1991) of 12 studies from 1938 to 1981 of the eutectic composition and temperature of the system. In this review no mention is made of the solid-liquid equilibrium described by Boisbaudran (Boisbaudran 1885), which might be assumed to be due to difficulties in obtaining sufficiently pure metals.

In summary, eutectic blends of gallium and indium have been described since the late 1800s, and have a unique property of being liquids at room temperature.

Flexible self-healing circuits

Self-healing materials are designed to repair and recover from damage without manual intervention. An early demonstration of what was then termed a self-repairing polymer was demonstrated by White et al. (White et al. 2001), where a ternary system contained emulsion of both catalyst and ‘healing agent’ (monomers) within a polymeric matrix. Crack formation would lead to rupture of healing agent microcapsules, which would then come in contact with exposed catalyst particles to polymerize within the crack. Recovery of up to 75% of tested mechanical properties was reported. This first paper on self-healing polymers has been followed by many similar investigations (Wu, Meure, and Solomon 2008).

Of particular interest are self-healing conductive polymer systems. These materials have applications in sensors, where damage to an overall structure could not only be repaired, but also be signaled by an integrated circuit. For example, Carlson et al. note that surface inspection is often done visually and can be impeded in situations where access is restricted (e.g. within the body of an aircraft) (Carlson, English, and Coe 2006). Addressing this problem, the authors design a self-healing polymer skin to be applied over a surface. The polymer uses encapsulated epoxy to self-heal punctures. Layers of copper within the polymer produce electromagnetic resonance when powered, and aberrations in the produced signal would signal to an operator that the surface has been damaged.

The approach followed by Carson et al. is somewhat limited in that, unlike the polymeric matrix, the copper circuit cannot self-repair; essentially the system constitutes an antenna which produces different signals based on its geometry. For more complex systems, it would be ideal if both circuit and matrix were capable of self-repair. The key limitation in self-repairing circuits is that traditional conductive materials, unlike viscoelastic polymers, are elastic materials that cannot reorient or reform after mechanical failure to reach and interact with a damage site. These materials would be generally limited to passive systems such as the one described.

A significant advance in self-healing conductive polymeric materials is described in a 2008 paper by Dickey et al. The authors describe previous systems where solder is injected into microchannels and permitted to solidify. This fabrication technique allows circuits to be constructed within a composite or polymeric matrix. Healing is then possible by reheating the solder to liquefy it and then allowing it to re-solidify. However, such a system cannot self-heal since heating is required. Additionally, the relatively high melting points of even low melting point solders (e.g. pure indium, 156.6 °C) limit the possible matrix materials.

To address this limitation, the authors propose the use of liquid metal alloys, such as eutectic gallium–indium (EGaIn), to fill microchannels in these structures. A property of EGaIn makes it particularly well-suited for this application: when exposed to air, a thin oxide film develops on the surface of the EGaIn structure (Xu et al. 2012). This film allows the formation of stable micro-scale structures (the authors demonstrate 1 μm cones). However, when this thin layer is disrupted, the metal reflows until the film is recovered.

To demonstrate the effect of the oxide skin on EGaIn structure, Dickey et al. created microchannels in polydimethylsiloxane (PDMS) varying from 40 to 20 μm in diameter, both untreated and treated with HCl to inhibit film formation. Experiments were performed both with EGaIn and with mercury, which is likewise liquid at room temperature but does not form an oxide layer.

While mercury and EGaIn + HCl were found to behave solely as Newtonian fluids, EGaIn showed an elastic response up to a critical surface stress of about 0.5 N/m, after which it flowed freely. This property, along with the ability to be worked at ambient temperatures, could permit many applications in organic and flexible electronics (e.g. sensors) and is promising for fabrication of self-healing conductive materials.

A series of practical demonstrations of self-healing circuits have followed the article by Dickey et al. In 2012, Blaiszik et al. encapsulated a thin conductive layer of metal with polymer (Blaiszik et al. 2012). The surface of the metal layer was deposited with small droplets of EGaIn. Breakage of the polymer and embedded thin film would rupture the EGaIn droplets, which would then flow into the crack, restoring the integrity of the circuit. Potential applications would be in improving redundancy for circuits exposed to stress.

A 2013 article by Palleau et al. demonstrates a self-healing wire that can heal both mechanically and electrically (Palleau et al. 2013). A cylindrical self-healing polymer (proprietary: Reverlink) encapsulates a channel filled with EGaIn. The polymer is elastic, and the liquid metal core deforms easily with the polymeric shell. Repair is demonstrated for complete rupture of the wire: the authors cut the wire with scissors, then hold both ends together to allow both the polymer and the metal to join again. Their system allowed for damage to be repaired within 10 minutes from rupture. Beyond elastic wires, applications could include reconfigurable circuits (cut and rejoin to re-route). This demonstration is exceptional in that it allows for healing of both polymer and conductor – articles described subsequently will be generally limited to creating elastic self-healing circuits.

A similar concept to the self-healing wire was demonstrated by Tabatabai et al. where a flexible circuit is constructed layer by layer: first elastomer, then a circuit is printed with EGaIn, then the whole covered by elastomer (Tabatabai et al. 2013). This elastic circuit is essentially a more complex imagination of the self-healing wire, but without a self-healing polymer. This is one of the earlier papers by the group of Carmel Majidi (one of the main authors on the paper by Yan et al. (Yan et al. 2019)) on using EGaIn for flexible (and later self-healing) circuits. It follows two applications-oriented papers: one for pressure sensing with elastic circuits (potential application: active orthotics) (Park et al. 2010), one for flexible curvature sensors (application: monitoring joint motion) (Kramer et al. 2011). A similar paper by the same group demonstrates masked deposition to create elastic EGaIn circuits (Kramer, Majidi, and Wood 2013).

A new approach to EGaIn circuits is described by the same group in a 2015 paper by Fassler and Majidi (Fassler and Majidi 2015). As the liquid metal conductor, galinstan, a eutectic alloy of gallium–indium–tin with a melting point of −19 °C, is used. The polymeric matrix is again PDMS. However, rather than printing a circuit layer-by-layer, the authors mix PDMS and galinstan manually in a 1:1 volume ratio with a mortar and pestle before molding the solution and curing at 150 °C. The structure of the resulting mixture is a suspension of galinstan droplets (2–30 μm, verified by optical microscopy) within the elastomeric PDMS matrix. By application of localized pressure (1.7 MPa), galinstan droplets within the matrix can be ruptured, with subsequent fusion creating regions of conductivity. While the focus of this article is not self-healing, this approach is suggested as an advance for ease of fabrication. The authors note that these types of elastomeric conductive circuits would have applications in wearable and implantable medical devices where a circuit would need to deform with the body of the patient or user.

A follow-up article in Nature Materials shows some improvements from the 2015 paper, with interesting demonstrations of how this type of circuit (formed by localized pressure on a dispersion of EGaIn) could self-heal (Markvicka et al. 2018). The examples in this paper are impressive, including a four-channel serial clock display and a quadruped locomotor robot. Again, the elastomer used is PDMS, this time with EGaIn. Mixing in this study is performed using a planetary mixer, which gives particles of a similar size to those produced using a mortar and pestle (on the order of 50 μm). The images of droplets provided by the authors show that droplets range from about 25 – 75 μm, are mostly ellipsoidal, and are closely distributed in the matrix (loading was from 20 – 50% by volume). However, the particle size is quite large, and the method used to determine the particle distribution is not specified.

The circuits, however, do demonstrate self-healing in response to punctures. The authors use a hole punch to remove a segment if the circuit. The localized pressure, however, fuses the EGaIn droplets surrounding the hole, reconnecting the circuit. This impressive demonstration shows how a dispersion of EGaIn droplets in an elastomer can function as a self-healing circuit. In considering applications in wearable sensors, the authors note that the required pressure to create circuits (about 1.7 MPa) is significantly higher than loads that might be expected in this application (example: highest localized pressure under the foot of a walking adult: 0.33 MPa).

In summary, self-healing, flexible electrical circuits can be created in elastomers. The method used by Majidi et al. involves fabrication by creating a dispersion of EGaIn particles within an elastomeric matrix, then fusing the droplets by localized pressure. These circuits are both flexible and self-healing. However, the large droplet size is a limitation, and the morphology of the dispersion is not fully described in the articles considered. It is thus unclear to what extent this method may be practical. These shortcomings will be addressed by Yan et al. (Yan et al. 2019).

Atom transfer radical polymerization

Radical polymerization mechanisms produce products quickly, but control of molecular weight is often difficult (Painter 1997). One of the most reliable methods for producing polymers with excellent molecular weight control (very low polydispersity) is anionic polymerization, first described by Szwarc in 1956 (Szwarc, Levy, and Milkovich 1956). In this technique, an anion (typically a carbon anion) is formed at the the end of a growing polymer chain and attracts monomers. With pure reactants and solvents, these growing chains do not undergo termination reactions, allowing chains to be formed of equal length (assuming equal reactivity per chain). Block copolymers can also be formed by addition of a second monomer type when the initial monomer is exhausted. Despite these advantages, however, anionic polymerization is only appropriate for certain monomer types and requires carefully controlled conditions.

An important development aiming to combine the advantage of both anionic and radical polymerization strategies was the development of atom transfer radical polymerization (ATRP). This technique was first described by Kato et al. in 1995 (Kato et al. 1995). The concept proposed by these authors is fairly simple: in cationic polymerizations, carbocations are stabilized by certain covalent species, making these reactions possible. If radical polymerization sites could be stabilized, this could prevent termination and chain transfer (branching), allowing control of molecular weight. The authors devise a polymerization reaction for methyl methacrylate (MMA), initiated/stabilized by CCL₄, RuCl₂(PPh₃)₃, and MeAl(ODBP)₂. While the authors were unable to elucidate the role of each initiator/stabilizer, they confirmed that the polydispersity of produced polymer was low (M_w/M_n ≈ 1.4) and that the system allowed for a ‘living polymerization’ (more monomer could be added and continue to react when previous additions were exhausted). Hypothesizing based on the available rate evidence and theoretical chemistry that the mechanism was by radical addition, Kato et al. demonstrated a well-controlled living polymerization by a novel mechanism.

Wang and Matyjaszewski demonstrated in an article published later in 1995 (Wang and Matyjaszewski 1995) an ATRP reaction for polymerization of polystyrene with 1-chloro-1-phenylethane and CuCl/bipyridine. The reaction mixture was heated to initiate the reaction in a sealed glass tube under vacuum. Size-exclusion chromatography was performed to determine the molecular weight and polydispersity throughout the reaction. The linear increase in number average molecular weight was taken to indicate that the number of chains remained relatively constant throughout the reaction. The observed polydispersities (about 1.4) were lower than obtained by conventional radical polymerization (2 – 3). The authors speculate on other potential catalysts and coordinating species, many of which have been shown to be practical for ATRP: a radical polymerization with well-controlled molecular weight and the possibility to form co-polymers (Matyjaszewski 2012).

A significant body of work has been produced characterizing, extending, and applying ATRP, most of which is beyond the scope of this review. For the purposes of this review, one additional article will be considered, authored by the group of the corresponding author of the previous paper, Krzysztof Matyjaszewski (also co-author of the article by Yan et al. (Yan et al. 2019)).

In a 2017 paper published in Chemistry of Materials, Yan et al. describe the development of a tetherable ATRP initiator targeting metal oxides (Yan et al. 2017). The authors state that their goal is to support the development of polymer/metal hybrid materials. A polymer initiator that could be anchored to the surface of a metal would allow a variety of polymers and metals to be compatibilized.

Building on a review of common anchoring groups, the authors select a carboxylic acid group to interact with the metal surface. A short saturated carbon chain (11 C) connects the anchor to the catalytic site: a 2-bromoisobutyrate group. The mechanism of the reaction is a supplemental activator and reducing agent (SARA) ATRP reaction (Matyjaszewski et al. 1997), which requires a zero-valent metal catalyst (in this study, copper). The initiator is synthesized in a one-step reaction of ω-aminolauric acid with 2-bromoisobutyryl bromide. The initiator, 12-(2-bromoisobutyramido)-dodecanoic acid is referred to as BiBADA by the authors.

Experiments were performed on 17 metal oxide nanoparticle surfaces, of which surface ATRP reactions had not been previously reported for 10 by means of other tetherable initiators. The molecular weight and polydispersities obtained ranged from 10⁴ – 10⁵ and from 1.2 – 2.4, respectively. The authors hypothesize various potential causes for the somewhat high polydispersities, principally that the 2-bromoisobutyrate initiating group may limit initiation efficiency. However, BiBADA is shown to be useful due to its compatibility with a wide range of metal surfaces, potentially for grafting with a variety of polymers.

In summary, the development of ATRP has allowed for radical polymerizations with well-controlled molecular weight distributions. A tetherable initiator, BiBADA, can be used to initiate ATRP reactions from the surface of many metal oxides.

Liquid metal dispersions

In the articles earlier discussed involving circuits formed by localized pressure on dispersions of EGaIn in elastomer (Fassler and Majidi 2015) (Markvicka et al. 2018), two important limitations went unmentioned; namely, that mechanical mixing of EGaIn with polymer often leads to anisotropically distributed properties, and that micrometer-scale droplets are both larger than ideal as well as frequently irregularly shaped (Yan et al. 2019). Both issues are linked to the interfacial interaction of the metal and polymer, and work has been done in parallel to applications-focused research on compatibilizing these blends.

In a 2011 study, Hohman et al. describe a method for creating sub-100 nm dispersions of EGaIn particles in ethanol (Hohman et al. 2011). Hypothesizing that the EGaIn surface oxide layer increases the surface energy at the interface, stabilizing larger particles, a thiolic antioxidant is added to a solution of EGaIn in ethanol. Cycles of ultrasonication, filtration, and rinsing are capable of creating spherical sub-100 nm dispersions. Smaller particle sizes are demonstrated for longer sonication times and more cycles, but the authors note that the morphology is less regular. No mention is made of the stability of these dispersions over time. It is also unclear what effect the thiolic additive could have for applications involving dispersion of these EGaIn particles in polymer.

A similar approach was attempted by Farrell and described in a 2017 article (Farrell and Tabor 2017). EGaIn was this time suspended in chlorobenzene (under inert atmosphere, to limit oxide formation), and subjected to cycles of ultrasonication and rinsing. Unlike in the previous article, a detailed methodology was provided describing the determination of the particle size distributions. Particle sizes and oxide layer thickness were studied by scanning transmission electron microscopy and by X-ray photoelectron spectroscopy, and studied over time. The interest of the authors was principally the progression of oxide formation rather than the final particle size distribution, and it was found that addition of thiols was able to reduce the thickness of the oxide layer by about 30%. Particles were mostly less than 100 nm in diameter. Again, similar concern apply about the limitations of this method.

An improved methodology was used by Finkenauer et al. to measure both the particle size distributions and the dispersion yield for a variety of surfactants (Finkenauer et al. 2017). Using dynamic light scattering and UV spectroscopy, an optical method was developed to estimate yield, i.e. the quantity of EGaIn suspended within the solution versus the quantity added initially. The authors find that all systems, including pure ethanol + EGaIn, were able to create nano-scale droplet suspensions by the ultrasonication and rinsing method. However, of the eight surfactants tested (four thiol, two amine, 2 carboxylic acid), only two (mercapto-dodecane and mercapto-octadecane) were capable of statically exceeding the yield of the control (pure ethanol). These results suggest that using thiolated surfactants may improve both particle size and yield, but calls into question earlier results that did not consider yield.

In summary, attempts have been made to reduce the droplet size of EGaIn particles, focusing on reducing the integrity of the surface oxide coating. This work has typically been performed in solution and is of limited applicability to metal-polymer systems.

“Solution processable liquid metal nanodroplets by surface-initiated atom transfer radical polymerization"

Having now considered the applications of EGaIn alloys in self-healing circuits, the development of tetherable ATRP initiators, as well as the challenges of improving the morphology and anisotropy of EGaIn dispersions, the key article of this review can be considered.

The 2019 paper by Yan et al. (Yan et al. 2019) is a collaboration between several authors previously mentioned: Carmel Majidi (EGain self-healing circuits), Krzysztof Matyjaszewski (ATRP), as well as the first author, Jiajun Yan (BiBADA tetherable initiator). The collaboration is an application of the ATRP tetherable initiator to improve the dispersion of EGaIn in the self-healing circuits.

Early in the paper, the authors admit to the difficulties in morphology and isotropy of mechanically-mixed EGaIn dispersions. The primary goal of the paper is to compatibilize the metal-polymer interfaces to improve the properties of metal-polymer blends (thermal, mechanical, electrical, optical).

A one-pot reaction of BiBADA, EGaIn, and monomer is described to form EGaIn particles with tethered poly(methyl methacrylate) (PMMA), poly(n-butyl acrylate) (PBMA), poly((2-dimethylamino)ethyl methacrylate) (PDMAEMA) and poly(n-butyl acrylate-block-methyl methacrylate) (PBA-b-PMMA). The solvent was tetrahydrofuran (THF), and copper was used as a catalyst.

The EGaIn particles with tethered polymer were found to be stable in solution (THF) at concentrations of up to 1.5 mM EGaIn, while small-molecule surfactant systems described previously are admitted to show coalescence and precipitation within 24 hours for solutions of as low as 0.2 mM metal. However, there is no discussion of suspension yield, shown by Finkenauer et al. to be an important parameter (Finkenauer et al. 2017). Conversely, it may be noted that Finkenauer does not address stability (particle size distribution) over time.

Particle sizes were found to be in the range of 200 nm (PMMA) and 120 nm (PBMA), though a particle size distribution is not presented. The methodology for estimating the particle sizes is not provided and appears to be based on visual inspection of TEM images. Based on the provided images, the particles appear to be roughly spherical in THF solution, but more irregularly shaped when cast. In any case, the magnitude of the particle sizes is significantly better than the 50 μm reported for mechanical mixing (Markvicka et al. 2018).

Solubility of compatibilized EGaIn was sufficiently high as to permit casting directly from the THF solutions for PMMA and PBMA. These samples had EGaIn mass fractions as high as 30% (about 6 volume-%). Tensile tests of cast PMMA- and PBMA-EGaIn, as well as blends of EGaIn–PBMA with an elastomer (Sylgard 184), showed properties judged by the authors to be comparable to reference samples, suggesting that the liquid metal inclusions did not significantly (or at least adversely) affect mechanical properties.

To further demonstrate the versatility of their method, the authors also use the ATRP reaction to produce a PBA-b-PMMA-compatibilized EGaIn, again directly cast from THF to produce samples for tensile testing. The PBA-b-PMMA copolymer was selected as a known thermoplastic elastomer capable of being synthesized with the ATRP system (Dufour et al. 2008). This experiment demonstrated the possibility of directly casting a compatibilized metal-elastomer blend. The authors also test the thermal transitions of this material, and find that the melting point of the EGaIn inclusions is depressed from about 16 °C to −85 °C, consistent with previous work that they cite on size-dependent crystallization of metal nanoparticles. However, the authors do not note that there could be a contribution from the particular properties of EGaIn, which has been previously noted as amenable to super-cooling (Anderson and Ansara 1991). Observation of thermal transitions appears not to have been performed in the other polymer-metal blends.

Two important elements are missing in this paper and would thus constitute appropriate future work: First, while the bulk of the article deals with the mechanical properties of compatibilized polymer-EGaIn blends, no comparison is available to mechanically-mixed blends. If compatibilization is expected to improve these properties, it is essential to have a basis for comparison. The finding that the properties are not adversely affected in compatibilized blends implies that they are meaningfully worse in mechanical blends. It seems likely that, as for difficulties with particle size, morphology, and anisotropy, the mechanical properties of materials described in previous studies may have been optimistically omitted.

The second basis for comparison of compatibilized and mechanically-mixed EGaIn materials regards their thermal and conductive properties. These have been well-characterized for previous mechanically-blended materials, but are nowhere explored in this article. Again, it is not clear that the compatibilized blends are better conductors of heat or electricity. Similarly, the authors’ mention of challenges with anisotropy for mechanically-mixed blends is provided without any available characterization of these challenges.

Thus, there is not enough information available on the shortcomings of mechanically-blended materials to compare them to compatibilized blends. It is likewise hoped that future publications will provide more information on the applications performance of the compatibilized blends so they can be meaningfully compared to previous work.

Conclusion

Eutectic gallium-indium (EGaIn) alloys, liquids at ambient conditions, are promising materials for flexible and self-healing circuits. Self-healing is possible through their non-Newtonian fluid properties, due to the formation of an elastic oxide surface coating which allows for stability at rest and reflowing under stress. Flexible, self-healing circuits have been demonstrated for EGaIn-polymer systems, including for dispersions of EGaIn in polymer matrices. However, these dispersions have been produced by mechanical mixing, giving large particle sizes and anisotropic properties that remain challenges for the development of these functionalized materials.

Atom transfer radical polymerization (ATRP) is a powerful method for controlled free-radical polymerization reactions. Recent developments have included the development of an ATRP initiator that can be tethered to metal oxide surfaces.

Yan et al. (Yan et al. 2019) apply a tethered ATRP initiator to an EGaIn system to produce compatibilized EGaIn-polymer blends. These blends have good mechanical and dispersion properties, but have yet to be fully characterized for their thermal and electrical conductivity. Comparison to previous work with EGaIn dispersions will clarify the significance and relevance of this achievement to potential applications, particularly in elastic electronics such as wearables and biomedical implants.

Written for the course Systèmes polymères multiphases (GCH6108) at Polytechnique Montréal, December 2019