The synthesis of conductive nanoscale and mesoscale structures is essential for developing low-cost, energy efficient, flexible and lightweight devices for next-generation micro and nanoelectronics.1 Despite the wide variety of electronic materials available including soft organic materials such as conducting polymers, carbon nanotubes, as well as inorganic materials such as metals and semiconductors, the creation of high-conductivity and high-resolution microcircuits in high aspect ratio layouts is a major challenge.1 Conventional electronics is based on solid inorganic materials with very high conductivities (∼103–105 S/cm) but with limited mechanical robustness and flexibility.1b Moreover, the substrates used for flexible electronics are also not compatible with the high temperature processing generally needed for forming metal circuits. Recent attention has therefore focused on using organic materials such as polymers and carbon-based materials as the conduction medium. In the field of organic electronics, however, typical conductivities of soft organic materials are ∼10−6 S/cm, limiting their practical implementation in electronics.1f The development of hybrid organic-inorganic mesoscale structures is one key to realizing flexible and conductive microelectronics.2 Applications that could benefit from this development include flexible displays, solar cell metallization, radio frequency identification tags, and micro-antennas.
Traditionally, nanoparticle (NP) assembly is carried out either with top-down or bottom-up approaches.1b Top-down approaches involve depositing, patterning, and etching material layers. These invasive procedures typically rely on control of damage, and as the structures approach smaller length scales, the increased number of manufacturing defects makes device operation problematic. Of bottom-up techniques, printing is the most widespread method.3 However, printing cannot fabricate effectively at small length scales without special surface templating.4 Other bottom-up methods include trapping individual NPs5 and nanowires using applied electromagnetic fields,6 or ligands.7 Post-synthesis assembly techniques for integration into high-density device assemblies include electric and magnetic-field-assisted alignment,6 optical8 and optoelectronic tweezers,9 micro-fluidic flow channels10 and micro contact printing.11 These methods involve multiple steps and can be limited by low deposition rates, lack of permanent bonding mechanism, and low electrical conductivity of the resulting microcircuits.
Here, we report the directed, colloidal synthesis of conductive organic-inorganic hybrid mesoscale structures. The technique is simple but allows hierarchical assembly and patterning of hybrid materials. We use a focused laser spot to direct colloidal assembly of nanoparticles (NPs) into electrically conductive organic-inorganic hybrid mesoscale filaments with arbitrary permanent patterns on a glass surface. High aspect ratio structures are fabricated with high deposition rate from colloids containing carbon and gold NPs, as well as other NP types. Growth mechanisms elucidated through finite element and kinetic Monte Carlo calculations suggest that this optically directed mesoscale assembly and patterning (ODMAP) operates through optical trapping, convective fluid flow, and chemical interactions forcing NPs to fuse near the laser focus and to continuously grow as the spot is moved along the glass-colloid interface. The synthesized filaments exhibit high ohmic conductivities, ∼430 S/cm, which is supported by non-equilibrium Green's function based density functional theory calculations. As ODMAP is applicable to other types of NPs and can be integrated with existing fabrication processes, it represents a simple and economical route to realize a variety of large-area electronic devices.
The ODMAP process is a significant development over existing assembly methods including the previously demonstrated optically directed assembly (ODA) process which takes place near the surface/edges of an evaporating colloidal droplet.12 ODMAP differs from ODA in four aspects: (1) It doesn't require non-equilibrium conditions such as evaporation near a droplet surface. (2) It requires roughly 10-fold increases in optical intensity and/or NP concentration, although these are still relatively moderate and easily achievable. (3) It gives rise to NP assemblies physically bound to glass. (4) It is potentially much more useful due to the ability to readily deposit intricate 2D mesoscopic patterns on glass at high throughput and low cost.
Figure 1a is a schematic of ODMAP. Positioning a weak optical trap (10–25 mW, 0.3 μm2 spot size, wavelength 488 nm) at the colloid-glass interface of an aqueous colloidal droplet containing carbon and gold NPs leads to fusion of NPs within the focus, with the NP assembly securely attached to the glass surface. (Owing to the mild experimental conditions, the binding to the glass is likely due to van der Waals interactions, although further work is required to confirm this.) Moving the optical trap laterally allows complex 2D patterns to be drawn. (3D formations are possible as well, since the trap can be moved vertically above the glass surface). See Figure 1b–e and the video ODMAPV1.mov (Supporting Information (SI)). Depending on the scan speed, pattern widths typically vary from 0.5 to 4 μm. ODMAP at the colloid/glass interface displays a clear threshold in optical intensity that depends on the local opacity at the glass-water interface. At optical power levels in the range 10–25 mW, ODMAP typically requires initiation (e.g., by positioning the laser spot on an opaque cluster of NPs on the glass surface). The assemblies can be post-processed into more complex compositions. For example, one can rinse and blow dry the sample and then add more colloid or a different colloid and carry out further ODMAP.
Once initiated, ODMAP is self-sustaining and leads to the formation of high quality and high-resolution traces built from hybrid organic-inorganic materials. The procedure is also easy and fast: the trace “Hi!” in Figure 1 was made in less than a minute of real time. A resolution (width) of 3 μm is seen for in the “hand-written” traces of Figure 1b–e using ODMAP with a laser power of 20 mW. (Depending on write speed, smaller, possibly sub-micron, widths might be achieved.) While 3 μm resolution is very crude in comparison with, e.g., state-of-the-art semiconductor technologies, it is superior to printing techniques for synthesizing flexible organic circuits which are currently limited to resolutions >15–30 μm.14 Also note that in the field of organic/hybrid electronics, the lowest resolutions that can be attained using laser-curing approaches are ∼2–5 μm,15 which is comparable to what ODMAP can attain.14 In principle, the current ODMAP process can attain resolutions as low as 0.5 μm, which is equal to the width of the laser beam. This is already significantly better than conventional printing processes which have a resolution of ∼30 μm.14
To understand ODMAP, we use computational fluid dynamics (CFD), coupled with energy balance equations incorporating a heat source to simulate the heating of a 500nm carbon NP in a colloid. The optical trap acts as a heat source that induces hydrothermal convection within the colloid owing to buoyancy effects. Our CFD simulations suggest that fluid velocities can range up to 1000–2200 μm/s (Figure S1). Such fluid flow can concentrate the particles near the optical trap. The temperature distribution (SI, Figure S2) indicates significant heating near the laser focus. Temperature is maximum near the NP surface and rapidly falls further into the fluid (Figure S2). For laser power ∼20 mW, the temperature rises ∼40–45 K above room temperature. This rise is associated with a change in fluid density, which leads to buoyancy driven fluid flow. Gravity causes upward movement of the lighter fluid while the heavier fluid flows downward. The hydrothermal toroidal convection current acts in conjunction with the optical trap allowing us to pattern arbitrary mesoscopic 2D structures on a glass substrate. Experimentally, for a 50:50 carbon:gold colloid, we find an ODMAP deposition rate of ∼2000 μm3/s at 20mW and total particle concentration 1 particle/μm3. A kinetic Monte Carlo (KMC) simulation of ODMAP carried out with convection velocities from the CFD simulations yields a deposition rate of ∼1000 μm3/s, in reasonable agreement with experiment (SI, Figure S4). KMC simulations indicate that deposition rates are higher for higher carbon fraction and scale linearly with optical intensity. For filaments with 6 μm diameter, we estimate write speeds of 70 μm/s are therefore possible. For comparison, direct-write electron beam lithography has write speeds on the order of 1–10 nm/s, although it can generate much finer scale structures.16
The conductivity of the hybrid mesoscale filaments is evaluated using two-probe measurements, Figures 2a–d. The current-voltage (I-V) characteristics shown in Figure 2d are measured for a filament with length L ∼ 1.1 mm and cross sectional area A ∼ 26 μm2. The I-V characteristics are ohmic, i.e., V = IR, with the inferred resistance, R, indicated in the figure. The conductivity is (1/R)(L/A) ∼430 S/cm, several orders of magnitude higher than those reported to-date for organic nanoelectronics devices.1d, 7, 13 The best-known elastomers display a conductivity of 0.1 S/cm whereas metal and carbon-particle composites as well as carbon nanotube (CNT) based conducting materials have conductivities in the 10−3 to 50 S/cm range, significantly lower than those obtained for ODMAP filaments.17
We also determined the electrical conductivity of the Au-C filaments synthesized via a “coffee-stain” mechanism18 involving natural drying (Figure 2c). The measured resistance for the coffee-stain Au-C filament is ∼12.5 kΩ whereas the same for smaller diameter ODMAP filaments was 1 kΩ. The conductivity of the coffee-stain structure with 20 μm thickness and 0.5mm length is 1.5 S/cm. This value is two orders of magnitude smaller than the ODMAP conductivity, reflecting, presumably, varying porosity and carbon content.
TEM images of the related ODA filaments and MD calculations suggest that the ODMAP filaments are composed of metal NPs surrounded by amorphous carbon acting as glue.17 AFM images (SI, Figure S3) suggest that the ODMAP filaments have somewhat rougher surface features than ODA ones, presumably due to differences in the convection patterns for the two processes: ODA causes the NPs to flow parallel to the filament axis, while ODMAP causes a perpendicular NP flow pattern. The elemental composition (carbon and gold) in the ODMAP and coffee-stain filaments was measured by energy-dispersive X-ray spectroscopy (EDX). SI, Figures S5a-d show SEM images and Figures S6–S7 show EDX composition mapping, respectively, of an ODMAP filament made from amorphous carbon and gold. Similar analysis was also done on the coffee-stain (SI, Figures S8–S10). We find that the filaments in both cases are predominantly composed of carbon with gold fraction varying from 8–15%. Both carbon and gold are uniformly distributed across the mesoscale filament (SI, Figure S7). ODA and ODMAP filaments, owing to their high amorphous carbon content, should have mechanical properties intermediate between amorphous carbon and gold. However, the degree of softness (e.g., Young's modulus) and potential utility for flexible electronics remain to be determined.
To evaluate microstructural effects on conductivity, we perform non-equilibrium Green's function (NEGF) calculations using density functional theory (DFT). The NEGF-DFT calculations are carried out for a Au-C-Au junction as a model for contacts within an ODMAP filament. In all the cases, I–V characteristics are ohmic, in agreement with experiment (Figure 2). Figure 3 shows conductivities as a function of amorphous carbon layer thickness. Experimental TEM images (SI, Figure S3a) show carbon gaps between gold NPs distributed in the range 2–10 nm with average separation of ∼4 nm. Interestingly, the experimental conductivity, 430 S/cm, is in reasonable accord with the theoretical result, 656 S/cm, at this separation. Figure 3 shows that varying the thickness of the carbon layer separating the gold NPs (e.g., by varying colloidal composition) from 4nm to 1nm leads to conductivities that are an order of magnitude higher than those currently reported.
While “optical tweezer” approaches that assemble nanoparticles into patterns have indeed been around a long time, it is important to realize that when the laser is turned off in such experiments, the assembly falls apart.8 In our approach, assembly formation is irreversible, making it much more relevant to materials and device fabrication. To the best of our knowledge, aside from our work, there is only one prior report where irreversible structures were formed.8 This latter work involved relatively long (2–15 μm) VO2 nanowire components, and resulted in relatively crude and simple bridges. They were unable to assemble nanometer-sized particles since the primary binding mechanism in their case is via van der Waals interactions.
ODMAP should also be differentiated from laser based curing approaches that rely on metal NP ink or organometallic ink, where the NPs are thermally grown in situ and aggregated by continuous heating.19 These laser based curing approaches demand a complex synthesis process; the organometallic compounds are first produced by chemical reaction of a metal oxide and a reducing agent followed by solvent evaporation of the metal/organic complexes. We note that since the organometallic ink is transparent, it cannot be directly subjected to laser direct curing process owing to the absence of solid particles to directly absorb laser energy. Therefore, this process requires a short thermal baking period to nucleate the metal NPs. Once the metal NPs are formed, a focused laser is irradiated to melt and transform the NPs into a continuous metal. Note that the process relies on the low melting temperature of NPs and hence works best for smaller sized particles below 10 nm diameter. Also, since this procedure is much more complex, it has been mostly shown to work with metals like Ag19 and to our knowledge, has not been applied to other materials such as carbon, silicon, tungsten etc. In our work, we achieve directed assembly of small sized colloidal particles (50 nm and less) without using any organometallic or metal NPs ink. This assembly procedure is significantly different from any of the previously demonstrated laser assembly process and allows for efficient assembly of a wide range of nanometer sized organic and inorganic materials.
ODMAP can be used to assemble other hybrid materials, although the complete space of possibilities has not been fully mapped out. A simple hypothesis is that for ODMAP to occur one needs a colloidal solution with NP types A, X1, X2, etc., where A is an absorber NP, i.e., one that efficiently heats up within the laser focus and Xn are other NP types added to achieve specific properties such as conductivity. (More than one absorber NP is also possible.) The absorber A heats up, wets and binds the other NP types, acting as the glue. Up to now, A has been an amorphous carbon NP and we have had gold as the other NP type (X). However, if instead of carbon NPs we use CNTs, also known to be good absorbers, we find that we can still create ODMAP filaments. Figures S11a-b show two probe measurements carried out on a ∼600 μm long CNT-Au wire with a thickness of ∼6 μm. As in the case of amorphous carbon, the measured I-V characteristics are ohmic. The conductivity of the ODMAP CNT-Au mesoscale filament is ∼280 S/m, lower but comparable to our results with amorphous carbon. Although compositional analysis of the ODMAP filaments made from CNTs and Au NPs suggests a higher fraction of gold in comparison to those made from amorphous carbon and Au NPs (compare SI, Figures S5–S7 with S12–S14), our analysis of the SEM images (SI, Figures S5, S8 and S12) reveals much higher porosity in the filaments made from Au NPs and CNTs compared to those made from Au NPs and amorphous carbon, which may account for the lower conductivity.
Consistent with the absorber hypothesis above, we have achieved ODMAP with several other organic-inorganic combinations: (C, Ag), (C, Ag2O), (C, TiO2), and (C, Cu2O). The case of two absorbers, and one non-absorber (C, Ag2O, Au), has also been demonstrated. (As noted above, “C” could be amorphous carbon nanoparticles or CNTs or other allotropes of carbon). It is worth stating that the absorber need not be organic. For example, we find that ODMAP works with silicon nanoparticles instead of carbon ones. We have furthermore carried out structural (SEM) and compositional (EDX) characterization of some of these hybrid filaments [(C, Ag2O, Au) and (Si, Au)]. The characterization details for some of these hybrid materials are included in the supplementary information (SI, Figures S15–S20).
Although the synthesis procedure demonstrated above allows for patterning on a solid substrate such as glass, we believe that the ODMAP process can also have an important role to play in the area of flexible electronics. In recent years, the synthesis and assembly of hybrid materials on a flexible substrate represents one of the critical barriers to develop flexible electronics: conventional fabrication methods work only with rigid wafer-based technologies.1b We have devised a simple post-synthesis method (inspired by the famous graphene “scotch tape” experiments20) for transferring our filament structures on to a flexible material such as a temperature resistant scotch tape. Figure 4 shows the optical image of a (Si, Au) filament transferred to temperature-resistant tape. We believe this is a significant result and that flexible electronics represents an important future direction to pursue with our approach. Apart from this transfer process, the ODMAP process might also be used to directly synthesize the filaments on a flexible substrate, which we plan to explore further in the future. Stretchable electronic materials such as those grown using the ODMAP process have potential as flexible sensor tapes for monitoring structural integrity, active antennas, printer cables, and flexible components for communications (cell phones), neural probes, biotic-abiotic interfaces, and painless drug delivery systems.1c
To summarize, we demonstrated a directed assembly phenomenon that allows for controlled synthesis of conducting 2D organic-inorganic hybrid mesoscale filament patterns on glass. This work represents a significant advance over other laser curing techniques considering that (a) arbitrary patterns can be formed with ease, (b) the process applies to a number of different organic and inorganic nanoparticle types, (c) the assembled mesoscale structures are irreversible, (d) the assembled structures can be electrically conducting and (e) a simple process allows for transfer of these filaments on to a flexible substrate. Moreover, our process is extremely cheap–only simple colloidal solutions and a very low power laser are involved. Thus ODMAP offers control, simplicity, high throughput, and possible cost advantages for NP assembly into functional devices. ODMAP could therefore open up an avenue for synthesis of a new class of hybrid organic-inorganic materials that will allow for the generation of “on-demand” and re-configurable circuitry.