Photothermal Actuation of Thick 3D-Printed Liquid Crystalline Elastomer Nanocomposites

Abstract

Liquid crystalline elastomers (LCEs) are stimuli-responsive materials that transduce an input energy into a mechanical response. LCE composites prepared with photothermal agents, such as nanoinclusions, are a means to realize wireless, remote, and local control of deformation with light. Amongst photothermal agents, gold nanorods (AuNRs) are highly efficient converters when the irradiation wavelength matches the longitudinal surface plasmon resonance (LSPR) of the AuNRs. However, AuNR aggregation broadens the LSPR which also reduces photothermal efficiency. Here, the surface chemistry of AuNRs is engineered via a well-controlled two-step ligand exchange with a monofunctional poly(ethylene glycol) (PEG) thiol that greatly improves the dispersion of AuNRs in LCEs. Accordingly, LCE-AuNR nanocomposites with very low PEG-AuNR content (0.01 wt%) prepared by 3D printing are shown to be highly efficient photothermal actuators with rapid response (>60 % strain s⁻¹) upon irradiation with near-infrared (NIR; 808nm) light. Because of the excellent dispersion of PEG-AuNR within the LCE, unabsorbed NIR light transmits through the nanocomposites and can actuate a series of samples. Further, the dispersion also allows for the optical deformation of millimeter-thick 3D printed structures without sacrificing actuation speed. The realization of well-dispersed nanoinclusions to maximize the stimulus-response of LCEs can benefit functional implementation in soft robotics or medical devices.

Graphical Abstract

A two-step process to functionalize gold nanorods (AuNRs) with poly(ethylene glycol) (PEG) thiol results in improved AuNR dispersion in liquid crystalline elastomers (LCEs). Improved dispersion and low PEG-AuNR loadings in 3D-printed LCE-AuNR enable efficient photothermal heating, rapid actuation with contractile strain rates exceeding 60 % s⁻¹, and complex photothermal deformation of ~mm thick 3D-printed structures.

Introduction

Nematic liquid crystalline elastomers (LCEs) are loosely crosslinked polymer networks that retain an average molecular orientational (nematic) order, which can be programmed by a variety of methods including surface alignment, optical patterning, and shear forces.^[1] Direct ink write (DIW) 3D printing has emerged as a powerful tool for additive manufacturing of LCEs with spatial control over molecular order, as the shear and extensional forces of extrusion align mesogens with the print path.^[2,3] Upon application of a stimulus, typically in the form of heat, nematic LCEs undergo an order-disorder transition resulting in macroscopic contraction along the direction of molecular alignment and expansion in orthogonal axes.^[4] Upon removal of the stimulus (e.g., cooling), LCEs recover their original geometry as permanent crosslinks enforce the original molecular order in the network.

This two-way shape memory has unlocked applications in soft robotics, artificial muscles, and even medical devices.^[5,6] However, the direct application of heat to drive the underlying phase transition is subject to the limits of heat transfer, constraining the rapid actuation of thick LCEs. Electrothermal actuation via Joule heating has been demonstrated by incorporating liquid metal either directly into the network or via a core-shell 3D printing strategy, but these mechanisms are still slow and require direct connection to a voltage source.^[7,8] Light is an ideal stimulus for LCEs as it enables remote actuation with spatiotemporal control.^[9] The use of azobenzene-containing LCEs for UV-light-driven actuation is well studied, but remote actuation of thick samples is impossible due to the high absorbance of azobenzene monomers.^[10] Furthermore, additional stimulus in the form of visible light is required for reversal of UV-induced azobenzene isomerization, which otherwise proceeds through thermal relaxation and can take hours or even days in the case of more stable azobenzene derivatives.^[11]

To address some of these limitations, photothermal heating has been employed by incorporating AuNRs, carbon black, carbon nanotubes, or even mesogenic chromophores into the LCE network.^[12] These photothermal agents convert optical energy into heat from within the LCE matrix itself, resulting in rapid heating and high strain rates upon irradiation with light. AuNRs are ideal photothermal agents, attributable to a very high molar extinction coefficient and near-unity photothermal conversion efficiency.^[9] AuNRs convert optical energy to heat through a surface plasmon resonance (SPR) mechanism in which the conductive band electrons oscillate in resonance with a specific wavelength of light and generate heat via a series of photophysical processes.^[13] The longitudinal SPR (LSPR), dictated by the aspect ratio of the nanorod, is stronger than that of the transverse SPR and 5-6 orders of magnitude larger than that of dye molecules. Furthermore, while dyes suffer from photobleaching and degradation, plasmonic effects in AuNRs are extremely stable as long as thermal reshaping does not occur.^[14]

While AuNRs exhibit highly desirable photothermal properties in isolation, SPR coupling between two adjacent AuNRs results in blue or red shifting of the LSPR for side-to-side and end-to-end alignment, respectively.^[15] With larger aggregates, effects on LSPR are more complex but follow a similar trend typically accompanied by peak broadening.^[16-18] In the context of LCE-AuNR nanocomposites, nanoparticle aggregation impairs local mesogen alignment and exacerbates nanointerfacial stresses that develop under strain.^[19] Thus, AuNR dispersion is critical to both the photothermal and thermomechanical properties of the resulting nanocomposite. AuNRs are typically generated via a seed-mediated process in the presence of the charged surfactant cetyltrimethylammonium bromide (CTAB), which forms a stable bilayer on the surface of the AuNR to prevent aggregation in aqueous solution.^[20] However, the presence of this charged molecule prevents the transfer and dispersion of AuNRs into organic solvents, which is necessary for subsequent mixing with liquid crystalline (LC) prepolymers.^[21] Therefore, the replacement of CTAB with a different surface ligand is necessary for relatively homogenous incorporation into LCEs.

Generally, techniques for ligand exchange harness the high affinity of thiols for Au, using poly(ethylene glycol) (PEG) or alkyl thiols to displace CTAB and form a covalent brush layer that limits AuNR aggregation and phase separation from liquid crystalline mesogens.^[22] Several prior attempts to disperse AuNRs in LCEs for photothermal actuation yielded partial dispersion.^[23,24] For instance, Hauser et al. used a one-step, one-pot ligand exchange with 2 mM PEG-thiol (PEG-SH; 800 g mol⁻¹) to functionalize AuNRs in water before mixing with LC monomers and photopolymerizing the thin (30 μm) films.^[25] While the resulting LCE-AuNR films exhibited high optical transparence and were capable of developing in-plane strains evenly throughout the thickness, TEM images demonstrated heterogeneously distributed PEG-AuNRs (primarily surface segregation) and some limited AuNR clustering.^[25] Wang et al. used the same approach to create thicker (100 μm) LCE-AuNR for photothermal actuation, but these nanocomposites exhibited bending upon NIR irradiation due to the development of a temperature gradient through thickness of the material.^[26] This was likely a result of AuNR aggregation causing strong absorbance and scattering at the irradiation front. These one-step ligand exchanges may have suffered from the preferential binding of thiols to the poles of the AuNRs, which have a lower CTAB packing density.^[15] Then, upon removal of CTAB, side-by-side dimerization and further aggregation occurs as the unfunctionalized sides of nanoparticles come into contact and irreversibly anneal. The effects of AuNR aggregation via this mechanism restrict the application of these photothermally-driven LCE actuators to thin samples incapable of substantial work output and conventional LCE actuation modalities.

Here, we use a two-step ligand exchange strategy with PEG-SH that results in PEG-AuNRs capable of maintaining long-term dispersion in organic solvents (e.g., DCM). After combining PEG-AuNRs and LC oligomers in DCM, removal of solvent results in near-uniform dispersion of PEG-AuNRs in a 3D-printable LC nanocomposite ink. By varying the concentration of PEG-AuNRs and the intensity of light, we could tune both the equilibrium temperature and the rate of temperature change in the LCE-AuNRs. Using a low concentration of PEG-AuNRs, 3D-printed LCE-AuNR rectangular actuators achieved full-thickness (250 μm) photothermal heating rates up to 150 °C s⁻¹, resulting in characteristic LCE directional contraction with rapid rates exceeding 60 % s⁻¹. Furthermore, NIR light was found to partially transmit through 250 μm thick 3D-printed LCE-AuNR linear actuators, allowing for photothermal actuation of a series of spatially separated elements. Finally, we demonstrate photothermally driven complex deformation of multilayer 3D-printed +1 disclinations reaching 1 mm in thickness.

Results and Discussion

Two-step ligand exchange enhances PEG-AuNR dispersion in LCEs

AuNRs with a 4:1 aspect ratio (48 x 12 nm), corresponding to an LSPR of 788 nm, were prepared by a concentrated seed-mediated synthesis (Figures S1, S2A).^[20] AuNRs were subsequently functionalized with an 800 g mol⁻¹ poly(ethylene glycol) thiol (PEG-SH) via a two-step exchange process adapted from Kinnear et al. (Figure 1A; Supporting Information).^[27] UV-vis-NIR spectra and dynamic light scattering confirmed the dispersion of PEG-AuNRs in DCM, with a maxima in the size distribution of 50 nm and a small red shift in the LSPR to 818 nm due to the increased refractive index (RI) of DCM (1.42) compared to water (1.33) (Figures S1A, S1C, S2B).^[13,15,28] PEG-AuNRs were also prepared via a conventional one-step ligand exchange, which showed evidence of aggregation (broad, red shifted LSPR) after transfer to DCM (Figure S1C).

Figure 1. LCE-AuNR fabrication and characterization.

(A) Illustration of two-step ligand exchange process, chemical structure of monomeric precursors, and LCE synthesis to prepare LCE-AuNR via DIW 3D printing. (B) Thermal analysis of a polydomain LCE-AuNR film prepared with 0.01 wt% PEG-AuNR. (C) UV-vis-NIR spectra of polydomain LCE-AuNR with 0.01 wt% PEG-AuNR. The dashed line is at 808 nm, the NIR laser wavelength. (D) TEM image of polydomain LCE-AuNR with 0.1 wt% PEG-AuNR. Scale bar = 500 nm. (Inset) higher magnification image of boxed region. Scale bar = 100 nm.

LCE-AuNR were prepared using a conventional two-step polymerization reaction.^[3] In the first reaction, thiol-terminated LC oligomers were generated through a base-catalyzed thiol-Michael addition reaction between mesogenic diacrylates and a molar excess of a low-molecular weight dithiol (Figure 1A). C6BAPE was incorporated at a 1:4 molar ratio to C3M to lower the nematic to isotropic transition temperature (T_NI) and increase the strain response per unit temperature.^[29] Oligomers, crosslinker (GBDA), and photoinitiator (I-369) were dissolved in DCM, to which PEG-AuNRs in DCM were added at different concentrations based on their absorbance at 400 nm.^[30] DCM was removed via rotary evaporation and the nanocomposite ink was fully dried in a vacuum oven overnight (Figure S2C). Films from the LCE-AuNR mixture were prepared in the polydomain orientation via photopolymerization between Rain-X coated glass slides with 250 μm spacers at 65 °C (Figure S2D). The thermal properties of the polydomain LCE-AuNR sample were characterized by differential scanning calorimetry and typified by a glass transition temperature (T_g) of −15 °C and a nematic-isotropic transition temperature (T_NI) of 57.5 °C (Figure 1B).

AuNR dispersion in the LCE matrix was assessed with UV-vis-NIR spectroscopy and transmission electron microscopy (TEM). The UV-vis-NIR spectra of the polydomain LCE-AuNR and the PEG-AuNRs in DCM are very similar, exhibiting a narrow LSPR centered at 850 nm. This is strong evidence that the AuNRs maintained their dispersity after formulation and polymerization (Figure 1C). The red-shift in the LSPR is attributed to the increased refractive index of LCEs (~1.5–1.7).^[28,31] TEM images confirm that the PEG-AuNR were largely dispersed as single nanorods with a few nanorod pairs observed (Figure 1D). Furthermore, PEG-AuNRs were homogenously distributed throughout the thickness of the LCE with no evidence of surface segregation (Figure S3A). In some regions, small clusters of <20 PEG-AuNRs were observed (Figure S3B). Prior investigations identified trace amounts of CTAB present on PEG-AuNRs even after a two-step ligand exchange, likely intercalated within the PEG brush, which could result in limited segregation during solvent evaporation.^[27] For comparison, TEM of LCE-AuNR with PEG-AuNR prepared via a typical one-step ligand exchange exhibited large (>1 μm) PEG-AuNR aggregates with very few isolated PEG-AuNRs in the LCE matrix (Figure S4A). Thus, the two-step ligand exchange was found to significantly enhance PEG-AuNR dispersion both in organic solvent and in the LCE matrix.

Polydomain LCE-AuNR exhibit predictable and efficient photothermal response

Next, we assessed whether the improved PEG-AuNR dispersion enhanced the photothermal efficiency of these LCE-AuNR nanocomposites. Polydomain LCEs were subsequently cut into strips and placed into an optical enclosure with a thermal imaging camera. The temperature of a control sample without PEG-AuNRs increased by just 2 °C even when irradiated with relatively high intensity (5.85 W cm⁻²) NIR light. In contrast and as expected, the LCE-AuNR nanocomposite with 0.1 wt% PEG-AuNR exhibited rapid temperature increases upon NIR (808 nm) irradiation even at low intensities (Figure 2A). At 0.25 W cm⁻², the polydomain sample achieved a temperature of 100 °C in under 4 s and reached its equilibrium temperature of 117 °C within 9 s. Both the equilibrium temperature and the rate of temperature change of the LCE-AuNR (0.1 wt%) increased linearly with intensity over the range that was tested (Figure 2C). In comparison, we also assessed the photothermal properties of a polydomain LCE-AuNR sample containing 0.1 wt% PEG-AuNRs prepared via a one-step ligand exchange with PEG-SH. While the relationship between equilibrium temperature and intensity remained linear, significantly higher intensities were required to achieve the same equilibrium temperature and rate of temperature change (Figures 2C, S4B). These results confirm that improving AuNR dispersion in LCE-AuNR can significantly enhance photothermal efficiency.

Figure 2. Photothermal properties of polydomain LCE-AuNR.

(A, B) Polydomain LCE-AuNR temperature during irradiation with NIR laser (gray region) at various intensities with (A) 0.1 wt% and (B) 0.01 wt% PEG-AuNR. The control sample is the same LCE formulation without any AuNRs, irradiated at 5.85 W cm⁻². (C) Summary of polydomain LCE-AuNR photothermal performance. Equilibrium temperature (top) and maximum rate of temperature change (bottom) as a function of NIR laser intensity for different AuNR loading and ligand exchange strategies. Of note, the maximum temperature of the thermal camera is 160 °C (dashed line in top panel). Data plotted on this line is included for completeness, but the actual equilibrium temperature is greater than 160 °C.

We next considered whether we could reduce the PEG-AuNR concentration by an order of magnitude while retaining an efficient photothermal effect. Polydomain LCE-AuNR were prepared by the same process, but with addition of just 0.01 wt% PEG-AuNR. To achieve the same equilibrium temperature as the 0.1 wt% LCE-AuNR, the 0.01 wt% LCE-AuNR required ~10x the intensity of NIR irradiation (Figures 2B, 2C). Nevertheless, this sample exhibited the same linear increase of equilibrium temperature and maximum rate of temperature change with increasing NIR intensity across the range of intensities tested. Impressively, the photothermal efficiency of the LCE-AuNR with 0.01 wt% PEG-AuNR was estimated to be greater than 40%, without accounting for heat lost to the environment. Motivated by this high efficiency, subsequent examination in this report utilizes samples prepared with 0.01 wt% PEG-AuNR.

3D-printed LCE-AuNR actuate rapidly upon NIR irradiation

One of the limitations of the photothermal actuation of LCE-AuNR nanocomposites in prior reports is their tendency to absorb strongly at the irradiation front, resulting in a temperature gradient through the polymer. This thermal gradient results in bending upon irradiation rather than contraction.^[26] We hypothesized that using a low concentration of PEG-AuNR would reduce this gradient as light would be absorbed incrementally throughout the thickness of the sample. Thus, we used direct ink write (DIW) to 3-D print rectangular LCEs as linear actuators loaded with 0.01 wt% PEG-AuNR. Our formulation allowed for room temperature extrusion onto a room temperature substrate, which was intended to eliminate alignment gradients seen in other 3D-printed samples that require elevated extrusion temperatures.^[32] The resulting LCE actuators (250 μm thick) with PEG-AuNRs had a light brown tint compared to yellow coloration (associated with residual photoinitiator) observed in LCEs without AuNRs (Figures 3A, S2E, S2F). 3D-printed LCE-AuNR were birefringent when viewed with a polarized optical microscope, confirming their optical anisotropy after extrusion and photopolymerization (Figure 3B). Before assessing photothermal actuation, the thermotropic strain response of the 3D-printed LCE-AuNR was characterized by dynamic mechanical analysis (DMA). Upon heating from −20–100 °C, the LCE-AuNR actuators contracted by ~50 % with a peak derivative of −1.5 % strain °C⁻¹ at ~60 °C (Figure 3C). The strain response exhibits two separate slopes, potentially indicating a cybotactic nematic to nematic transition occurring from ~ 0–50 °C followed by a nematic to isotropic transition.^[33]

Figure 3. 3D-printed LCE-AuNR characterization.

(A) Images of 3D printed rectangular actuators with 0.01 wt% PEG-AuNR (left) and without PEG-AuNR (right). Scale bar = 5 mm. (B) Polarized optical micrographs of 3D printed LCE-AuNR at 45° (top) and 0° (bottom) offset to crossed polarizers. Scale bars = 400 μm. (C) Thermotropic strain behavior of 3D printed LCE-AuNR. (D) 3D-printed LCE-AuNR temperature (top) and strain (bottom) during irradiation with NIR laser (gray region) at various intensities. (E) Summary of 3D-printed LCE-AuNR equilibrium temperature and absolute value of equilibrium strain as a function of NIR laser intensity. (F) Maximum absolute value of strain rate plotted against maximum rate of temperature change. Marker color indicates the same NIR intensities shown in (D) from 0.14–5.65 W cm⁻², with the same color chosen for 11.29 and 19.76 W cm⁻². (G) LCE-AuNR actuation in 1-minute cycles (15 s irradiation, 45 s relaxation) at 5.10 W cm⁻². (H) Work capacity of LCE-AuNR as a function of load (Force) at various NIR intensities.

The actuators were then subjected to NIR irradiation, which demonstrated largely contractile actuation behavior, as well as some twisting motion due to small spatial variances in alignment that could be replicated by heating on a hot plate (Video S1). When a lightweight (1 g) binder clip was attached to the actuator to provide tension and stabilize the actuator, the twisting behavior disappeared (Video S1). This result suggests that the low concentration of PEG-AuNR enabled full-thickness heating and conventional LCE directional strain. Subsequently, the photothermal actuation performance was assessed at various NIR intensities. Samples were irradiated starting at 10 s and ending at 30 s after which the recovery was monitored (Figure 3D). As expected, increasing NIR intensity resulted in increased equilibrium temperature and strain (Figures 3D, 3E). As NIR intensity was increased from 0.1 to 1 W cm⁻², the equilibrium temperature rose above the T_NI and the equilibrium contractile strain increased from 6 % to 31 %. At NIR intensities above 1 W cm⁻², the equilibrium contractile strain reached a plateau value of ~40 %. Notably, the sample did not achieve the full contraction observed in Figure 3C (50 % strain) as the laser could not irradiate the entire sample length. While actuation strain and equilibrium temperature exhibited a nonlinear relationship, their derivatives with respect to time maintained a linear relationship with each other and with NIR intensity (Figures 3F, S5A). Remarkably, the contraction rate of LCE-AuNR exceeds 60 % strain s⁻¹ at a relatively high NIR intensity of 19.76 W cm⁻². The LCE-AuNR did not exhibit any evidence of a plateau in actuation rate within the range of NIR intensities tested suggesting that greater temperature and strain rates could be attained with a more powerful laser. Of note, focused sunlight was also capable of driving photothermal actuation, towards applications where laser irradiation might not be feasible (Video S2; Supporting Information).

While high intensities of NIR light enable rapid actuation of LCE-AuNR, the high temperatures may also result in LCE decomposition (Figure S5B) as well as photothermal reshaping of PEG-AuNRs, which affects their aspect ratio and thus their LSPR.^[14] For example, at 250 °C, AuNRs can thermally reshape into spherical Au nanoparticles within an hour. However, AuNRs were shown to withstand pulsed laser-induced temperatures of 750 °C due to rapid heat dissipation between pulses.^[14] Thus, photothermal reshaping of AuNRs is dependent both upon temperature and time. Furthermore, interfacial stresses between AuNRs and the LCE matrix may affect the recovery and/or strain response with repeated actuation cycles.^[19] Consequently, we assessed the stability of rapid, high intensity photothermal actuation by performing repeated actuation cycles with 1 min intervals. At 5.10 W cm⁻², the LCE-AuNR actuator achieved −40 % strain within 15 s and recovered back to 0 % strain within 45 s. This actuation stroke was cycled 19 times within a 20 min period with no change in peak strain or recovery, confirming the stability of our LCE-AuNR during rapid photothermally-driven actuation at high NIR intensities (Figure 3G). Furthermore, the relationships we have established between intensity and strain response may lead to applications in autonomous material systems, where a desired equilibrium strain can be reliably achieved with remote stimuli at fixed intensity.

To quantify work capacity and external efficiency of these 3D-printed LCE-AuNR, photothermal actuation was performed over a range of loadings and NIR intensity. LCE-AuNR actuators were able to lift ~2000 times their own weight when irradiated with 5.10 W cm⁻² NIR light (Video S3). Work capacity scaled with increasing load and NIR intensity, with a peak work capacity of 90 J kg⁻¹ at 38.3 W cm⁻² with a load of ~28 g (Figure 3H). However, the laser spot size used (5 mm diameter) limited the actuation strain to just −14%. If the spot size was increased, or multiple light sources were used, DMA results (Figure 3C) suggest a maximum strain of −50 %, corresponding to a theoretical work capacity of ~280 J kg⁻¹ that compares favorably with other photothermal LCE actuators. External efficiency, defined as work output divided by the input optical energy, also scaled with increasing load, but was negatively correlated with NIR intensity (Figure S5C; Supporting Information) Altogether, the enhanced AuNR dispersion in these 3D-printed LCE-AuNR enables high efficiency and work capacity, rapid actuation, and stability during repeated actuation cycles.

NIR light partially transmits through 3D-printed LCE-AuNR actuators

To verify that low PEG-AuNR loadings resulted in full-thickness heating in the 3D-printed LCE-AuNR, we next quantified the NIR absorbance of the actuator above its T_NI during irradiation (Figure 4A). The NIR laser was focused to a spot size smaller than the width of the 3D-printed actuator with a strip of polydomain LCE-AuNR behind it, such that any light that reached the polydomain sample had to first pass through the actuator (Figure 4A). At lower NIR intensities (≤ 1.59 W cm⁻²), the temperature profile in both materials was similar over time (Figure 4B, Video S4). While we expected a lower temperature for the polydomain LCE-AuNR given that the light is attenuated by the actuator, the polydomain material is opaque, which may improve local light delivery to PEG-AuNRs via scattering. Thus, the stronger absorbance of the polydomain material results in a system near equilibrium. Meanwhile, at higher intensities (>2 W cm⁻²) the temperature in the actuator was higher than that of the polydomain LCE-AuNR after just a few seconds of exposure (Figure 4B). As the temperature of the actuator reaches the T_NI and contracts along the nematic director, it also expands orthogonally, causing an increase in thickness. This action brings more PEG-AuNRs into the beam path, which then absorb more photons and generate additional heat. This phenomenon can be seen more clearly in Figure 4C, where the equilibrium temperature in the two samples is the same up to ~1.7 W cm⁻² (~60 °C) and then diverges as more NIR light is absorbed by the thicker actuator, leaving fewer photons for the polydomain sample to absorb.

Figure 4. LCE-AuNR (0.01 wt%) partially transmit NIR light.

(A) Setup for assessing NIR transmission, where the laser spot size (red aiming beam) is focused to be smaller than the width of the actuator in front of the polydomain LCE-AuNR. (B) Temperature of the actuator and polydomain LCE-AuNR during irradiation with NIR laser (gray region) at 1.59 (top) and 4.78 (bottom) W cm⁻². (C) Equilibrium temperature of LCE-AuNR actuator and polydomain in series, as well as the polydomain sample alone at various NIR intensities. The gray line indicates where the NIR absorbance of the LCE-AuNR actuator was calculated. (D) Photothermal actuation of three 3D-printed LCE-AuNR actuators in series. Top: before NIR irradiation, red dashed line indicates the top of all actuators, the green lines indicate the starting position of each actuator; pink arrow indicates the direction of the laser beam. Bottom: last frame of NIR irradiation; pink bars and arrows indicate the change in length of each actuator after irradiation. (E) Temperature (top) and strain (bottom) of all three actuators and the polydomain LCE-AuNR during NIR irradiation at 14 W cm⁻² (gray region). (F) Equilibrium temperature (top) and equilibrium strain (bottom) of all three actuators and polydomain LCE-AuNR at various NIR intensities.

Subsequently, the actuator was removed and the polydomain LCE-AuNR alone was irradiated at the same NIR intensities. Exploiting the linear relationship between NIR intensity and temperature in the polydomain samples, we calculated an absorbance value for the LCE-AuNR actuator (above T_NI) of 0.77 by comparing the intensities required to achieve the same temperature in the polydomain LCE-AuNR with and without the attenuating actuator in place. This partial transmission of NIR light through the actuator validates that using a low concentration of PEG-AuNR enables full-thickness heating and conventional actuation behavior in the LCE-AuNR nanocomposites.

Next, we explored the limits of remote photothermal actuation by irradiating multiple LCE-AuNR actuators in series (Figure 4D, Video S5). As before, the laser spot was focused such that its diameter was smaller than the width of the first 3D-printed actuator (near). Two more 3D-printed actuators were placed in series behind the first with 2.5 cm spacing, with a polydomain sample at the end to capture light that is transmitted through all three actuators. All four samples were monitored with thermal imaging and a video camera to enable quantification of the strain response. At 14 W cm⁻², the temperature of first two actuators increased above the T_NI, while the third (far) actuator and the polydomain sample demonstrated marginal increases in temperature (Figure 4E). The temperature of each actuator correlated with the strain, which reached −25 % for the first actuator and −15 % for the second. At higher intensities, the strain response of the middle actuator was faster with greater strain compared to the near actuator (Figure 4F). This observation is attributed to forward scattering from the first actuator, which resulted in a greater area of the middle actuator being irradiated (Figure 4D, Video S5). Photothermal actuation of multiple LCE in series, demonstrated here for the first time to our knowledge, offers new opportunities for multimodal actuation in applications where the use of multiple irradiation sources is not possible. For example, soft robots where actuators need to be tethered to the same location or overlap but extend in different directions to perform distinct functions simultaneously. Although we did not see evidence of PEG-AuNR alignment with the print direction, future studies might exploit polarization-specific absorption of AuNRs as an additional mechanism for enhanced control of photothermal actuation in series.

Complex photothermal deformation of thick 3D-printed LCE-AuNR

While these demonstrations prove the concept of photothermal actuation in series, light scattering suggested we were not taking full advantage of the optical energy employed. To gain an understanding of the scattering properties, we irradiated a 3D-printed actuator and placed a NIR laser visualization card behind it. With just 1 cm of separation, the scattered light illuminated a region of ~1 cm in width on the card, which increased to ~4 cm with 10 cm of separation (Figure S6). Therefore, we reasoned we could harness the full performance of these LCE-AuNR by photothermally actuating a thick LCE-AuNR where scattered and transmitted photons would be absorbed by PEG-AuNR deeper within the sample. As a final demonstration, we 3D-printed the LCE-AuNR ink into square +1 disclinations with varied thickness (1, 2, 4 layers; 250, 500, 1000 μm thick).

All samples, including the 1 mm thick sample, achieved their expected 2D to 3D deformation resulting in a pyramidal shape upon NIR irradiation at intensities as low as 3.18 W cm⁻² (Figure 5A, Videos S6-S8). This result confirmed that scattering was limiting the actuation performance of the three actuators in series, having a combined thickness of 750 μm. The strain was quantified over time using I-unit analysis, which we have previously implemented to measure actuation of +1 disclinations (Figures 5B, S7).^[34,35] Both the maximum rate of temperature change and the maximum strain rate increased with increasing NIR intensity, as was seen in the linear actuators (Figure 5C). While the single-layer sample exhibited the highest performance, the maximum rate of temperature change and strain rate of the 4-layer sample was unexpectedly higher than that of the 2-layer sample. Nevertheless, when comparing the timescales of actuation and relaxation, increasing sample thickness clearly results in slower actuation and relaxation (Figure 5D). At 12.74 W cm⁻², all samples reached 95% of their peak strain in under 20 s (Figure 5D), suggesting that it would be possible to print and remotely actuate even thicker 3D-printed constructs. Furthermore, as this demonstration required placing the 3D-printed construct on top of a glass coverslip, NIR light had to pass through the glass before reaching the sample. Thus, 3D-printed LCE-AuNR hold promise for applications where remote actuation is required due to the presence of obstructing, yet NIR-transmissive, materials such as glass or even biological tissues.^[6]

Figure 5. Complex photothermal deformation of thick 3D-printed LCE-AuNR (0.01 wt%).

(A) Images of fully deformed 3D printed +1 disclinations with 1, 2, and 4 layers, captured one frame after the NIR laser was turned off. Scale bars = 5 mm. (B) I-unit strain over time of 3D printed +1 disclinations with (top to bottom) 1, 2, and 4 layers during NIR irradiation (gray region). (C) Maximum strain rate plotted against maximum rate of temperature change during NIR irradiation, for 1, 2, and 4 layer +1 disclinations. The marker color indicates the NIR intensity. (D) Time for 1, 2, and 4 layer +1 disclinations to reach 95% peak actuation strain during NIR irradiation at 12.74 W cm⁻² (Actuation), and time to reach 105% of final strain (Relaxation).

Conclusions

4D printing of programmable materials responsive to spatiotemporal stimuli such as light afford unparalleled control over shape change. In this work, AuNRs were incorporated into 3D-printable LC ink, enabling both programmable shape change and remote control via the photothermal effect. AuNR dispersion was optimized through a two-step ligand exchange with PEG-SH, which allowed for low PEG-AuNR loading in LCEs without sacrificing photothermal performance. Importantly, very low concentrations of PEG-AuNRs eliminated photothermally induced temperature gradients typically observed in prior examinations of thick LCE-AuNR. As such, the LCE-AuNR nanocomposites maintain performance that are ideal for implementations in soft robotics, such as rapid actuation rates, high photothermal efficiency and work capacity, and stability over repeated actuation cycles. Furthermore, NIR light is partially transmitted through these 3D-printed LCE-AuNR, enabling complex actuation modes (e.g., actuation in series, 2D to 3D deformation of 1 mm thick 3D-printed structures). The remote, wireless, and local control of actuation demonstrated here may be particularly beneficial in implementations in atypical environments such as underwater, in space, and within the human body.

Experimental Section

Materials

Cetyltrimethylammonium bromide (CTAB) was purchased from GFS chemicals. HAuCl₄, AgNO₃, sodium borohydride, hydroquinone, GBDA (glyoxal bis(diallyl acetal)), base catalyst TEA (triethylamine), thermal inhibitor BHT (2,6-Di-tert-butyl-4-methylphenol), and mPEG-SH (methoxy PEG thiol; 800 g mol⁻¹) were purchased from Aldrich. BMEE (bis(2-mercaptoethyl) ether) was purchased from TCI Chemicals. C3M (1,4-bis(4-[3-acryloyloxybutyloxy] benzoyloxy)-2-methylbenzene) was purchased from Wilshire Technologies. C6BAPE (4-(6-(acryloyloxy)hexyloxy)phenyl-4-(6-(acryloyloxy)hexyloxy)benzoate) was purchased from Synthon. Photoinitiator I-369 (Irgacure 369) was obtained from IGM Resins. All chemicals were used as received. All solvents were reagent grade and purchased from Sigma Aldrich.

Synthesis of AuNRs

The Au seeds were prepared according to the typical synthetic route.^[20] CTAB (0.364 g) was added to 10 mL of 0.25 mM aqueous HAuCl₄ solution (solution A). The solution was briefly sonicated (30 s) and kept in a warm water bath (40 °C) for 5 min to completely dissolve CTAB and left at 25 °C for 10 min. A 0.01 M aqueous NaBH₄ solution was prepared and kept in an ice bath. NaBH₄ solution (0.6 mL, 0.01 M) was quickly added drop wise to solution A while stirring at 800 rpm causing the color of solution to become light brown as Au seeds form. Stirring continued for 1 min and the seeds were aged for 5 min. The AuNRs were prepared according to the scale up protocol.^[20] The growth solution was prepared by mixing CTAB (0.1 g), aqueous AgNO₃ solution (250 μL, 0.1 M), and aqueous HAuCl₄ solution (500 μL, 0.1 M). Note that the sequence of the additions is crucial. Next, aqueous hydroquinone solution (1.25 mL, 0.1 M) was added to the growth solution as a mild reducing agent. The resulting solution was vortexed until it turned a milky white color (solution B). Solution B (350 μL) was introduced into the seed solution. After 2 h, aliquots of solution B (500 μL each) were added four times at 20-minute intervals. The solution was left undisturbed for 2 h, leading to the formation of AuNRs with an aspect ratio of 4 (average diameter of 12 nm and length of 48 nm). The as-made AuNRs were centrifuged twice at 15,000 rpm for 20 min to remove excess reactants and redispersed in 0.01 M CTAB solution. Following the third centrifugation (15,000 rpm x 20 min), the supernatant was decanted, and the concentrated sediment was subsequently diluted for the ligand exchange process. Some of the AuNRs, having identical specifications, were provided by UES, Inc. (AURA Gold Nanoparticles).

Two-Step Ligand Exchange

Two hundred microliters of concentrated AuNRs (15 nM; 5 mM CTAB) were added to 800 μL of 3.75 mM mPEG-SH in sonicated diH₂O, resulting in the following concentrations: AuNR: 3 nM; CTAB: 1 mM; mPEG-SH: 3mM. The solution was placed on a rotating stand for 24 h. AuNRs were centrifuged at 15,000 rpm x 20 minutes, supernatant discarded, and AuNRs resuspended in a mixture of 15 mM mPEG-SH in 90% ethanol (v/v) in sonicated diH₂O. The solution was placed in a water bath sonicator at room temperature for > 2 h, centrifuged again, and supernatant discarded. AuNRs were resuspended in 1 mL DCM and transferred to a glass scintillation vial for storage.

One-Step Ligand Exchange

One milliliter of AuNRs (10 nM; 10 mM CTAB) was added to 9 mL of 5 mM mPEG-SH in diH₂O, vortexed, and placed on a rocker plate overnight at room temperature. AuNRs were centrifuged, supernatant discarded, and AuNRs resuspended in 2 mL ethanol, sonicated, and centrifuged again. The supernatant was discarded and the AuNRs resuspended in 2 mL DCM and transferred to a glass scintillation vial for storage.

AuNR Sizing and Zeta Potential Measurements

A Zetasizer Nano ZSP (Malvern Panalytical Inc.) instrument was used for all measurements. Sizing was performed in a low-volume 3 mm path length cuvette (Malvern Panalytical Inc.). Aliquots of AuNRs were taken at different steps of the two-step exchange process (before the first step, 24 h after incubation in 3 mM mPEG-SH, 2 h after incubation in 15 mM mPEG-SH, and after resuspension in DCM) for sizing analysis, with the refractive index in the SOP matched to the solvent in each step. Three technical replicates were performed and the average ± SD of the technical replicates plotted. For surface zeta potential, a disposable zeta cuvette (Malvern Panalytical Inc.) was used with AuNRs taken at the same timepoints. Three technical replicates were performed and the average with all three replicate values plotted. Measurements of the surface zeta potential of PEG-AuNRs in DCM was impossible due to bubble formation between the electrodes creating artifacts.

UV-Vis-NIR Measurements and Calculation of AuNR Concentration

An Agilent Cary 7000 instrument was used for all measurements. For AuNRs in solution, the same low-volume 3 mm path length cuvette was used. The instrument was zeroed and blank spectra of the solvent without the AuNRs were used for background subtraction. Solid samples were attached to black metal plates with a circular hole in the center for light to pass through. The plates were taped directly in front of the detection window to reduce scattering losses. A background of the same polymer formulation and thickness, without AuNRs, was used for background subtraction. Spectra were taken from 1400 – 375 nm with 1 nm steps. For calculation of AuNR concentration in solution, the optical density at 400 nm, with correction for the 3 mm cuvette path length, was used to convert concentration into mg mL⁻¹.^[30]

Synthesis of LCE-AuNR

C3M and C6BAPE were weighed out into a 20 mL scintillation vial at a molar ratio of 4:1 C3M:C6BAPE. BHT (1 wt%) was added as a thermal inhibitor. The solids were melted at 200 °F using a heat gun, with intermittent vortexing to ensure complete mixing, and placed on a hot plate set to 65 °C. BMEE was added at a 1:0.8 molar ratio (thiol:acrylate) and the mixture vortexed. 1 wt% TEA was added and the solution vortexed thoroughly. The vial was left on the hot plate at 65 °C for at least three hours. Thiol-terminated oligomers were weighed into separate 20 mL scintillation vials, to which DCM was added to dissolve the oligomers. The crosslinker GBDA was added at 0.2:1 molar ratio (ene:thiol). I-369 (1.5 wt%) was also added to the solution. For LCE containing PEG-AuNR, the volume of PEG-AuNR in DCM added to the solution was calculated as wt% based on the concentration of PEG-AuNR in mg mL⁻¹ and the weight of the oligomers plus the weight of GBDA. The solution was mixed by vortexing and placed on a rotary evaporator, with the vial sitting in a water bath sonicator set to room temperature. After most of the DCM had been removed, the vial was placed in a vacuum oven overnight at 65 °C to full dry the LC ink. To generate polydomain LCE (± PEG-AuNR), the LC ink was heated to 65 °C and scooped onto a Rain-X treated glass slide with 250 μm Teflon spacers. Another Rain-X treated glass slide was placed on top and the ink photopolymerized with UV light (365 nm; 50 mW cm⁻²) for 10 min.

3D Printing of LCE-AuNR

A Hyrel3D System 30M printer equipped with KR2 printhead and custom low-volume HTK print cartridges was used for all 3D printing. The LC ink (± PEG-AuNR) was heated to 65 °C, loaded into the print head, the plunger advanced until the ink was visible at the tip, and the print head heated upright at 65 °C for > 30 min to remove any air bubbles. The print head and ink were allowed to cool to room temperature before the 400 μm diameter nozzle was attached to the print head. The LC ink (± PEG-AuNR) was printed onto PVA-coated glass slides at room temperature, with a print speed of 6 mm s⁻¹, layer height of 250 μm, and linewidth of 400 μm. During printing, low-intensity UV-light (365 nm; 0.8 mW cm⁻²) was applied to lock in alignment, with full curing taking place after printing was complete using UV light (365 nm; 50 mW cm⁻²) for 10 min. The dimensions of each printed linear actuator were 20 mm x 5 mm. The dimensions of each printed +1 disclination were 10 mm x 10 mm.

Differential Scanning Calorimetry

A TA Instruments Discovery DSC 2500 was used for all measurements. Samples (5–10 mg) were loaded into TZero pans with Tzero hermetic lids, and subject to a heat-cool-heat cycle from −50–120 °C at a rate of 10 °C min⁻¹. Data was reported only from the second heating.

Thermogravimetric Analysis

A TA Instruments TGA 5500 was used. A 3D-printed LCE-AuNR (0.01 wt%) sample (12 mg) was loaded into a Platinum High Temperature pan and subject to a heat ramp from 20–500 °C at a rate of 20 °C min⁻¹.

Sectioning of LCE-AuNR and Transmission Electron Microscopy

Samples were either embedded in an epoxy (EPON-812; Electron Microscopy Sciences) and cut at room temperature on an ultramicrotome (Leica Microsystems EM UC6), or frozen with liquid nitrogen to −30 °C and cut on a cryo-ultramicrotome at −25 °C (Leica Microsystems EM UC6 ultramicrotome equipped with an FC6 cryochamber). Images were taken on a JOEL JEM-1400 (120 kV) or a FEI Tecnai T12 (120 kV).

Polarized Optical Microscopy

The molecular alignment of 3D-printed LCE-AuNR was confirmed by the presence of birefringence using a Nikon Eclipse Ci-Pol with a 5x objective. Samples were observed at 0° and 45° rotation between crossed polarizers.

NIR Irradiation and Imaging

Samples were placed in an optical enclosure and taped to various optical supports such that they were freely hanging in air. A NIR laser diode (808 ± 10 nm) coupled to a fiber optic with a collimator was used for irradiation (BWT Beijing Ltd.). The red aiming beam (635 ± 10 nm) was used to focus the collimator to the desired spot size, which was measured using calipers. Intensity (W cm⁻²) was calculated based on the laser power and spot size. A thermal imaging camera (FLIR A655sc; Teledyne FLIR LLC) and software (ResearchIR) were used to capture the temporal temperature change. Temperature measurements were taken at the center of the laser beam. In some experiments, a Canon 4k video camera was used to capture the actuation in real time.

Calculation of Photothermal Efficiency

A polydomain LCE-AuNR sample (0.01 wt% PEG-AuNR) with dimensions 0.45 x 0.95 cm was irradiated with 1.412 W cm⁻² using a laser spot size larger than the sample itself. The sample temperature increased from 24–84 °C in 20 s. The input energy (J) was calculated by multiplying the intensity, time, and sample area. The specific heat (J g⁻¹ °C⁻¹) of the sample was estimated using DSC, dividing heat flow (W/g; average between 24–84 °C) by heating rate (0.1667 °C s⁻¹). The thermal energy (J) required to heat the sample from 24–84 °C was calculated by multiplying the specific heat by the change in temperature (60 °C) and the mass of the sample (0.05 g). Photothermal Efficiency was calculated as the Thermal Energy required to heat from 24–84 °C divided by the Input Energy. Notably, this estimate did not account for heat lost to the environment during heating or any transmission of NIR light through the sample.

Actuation Testing

A TA Instruments 850 DMA was used for all measurements. For thermotropic actuation without NIR light, the 3D-printed linear actuator was secured between tensile clamps with a preload force of 0.001 N, cooled to −20 °C, and equilibrated for 5 min. The strain was measured during a temperature ramp from −20 – 100 °C at a rate of 3 °C min⁻¹. For photothermal actuation tests, the fiber optic was attached to the thermal camera and focused on the middle of the actuator, with a spot size of ~10 mm. A creep test was run with 0% stress to allow the instrument to measure strain during NIR irradiation without applying any force. Once the test was started, the thermal imaging camera was set to record, and the laser turned on. During data processing, the onset of irradiation was set to 10 seconds and the time values of the strain measurements were adjusted accordingly.

Calculation of NIR Absorbance of 3D-Printed LCE-AuNR

The equilibrium temperature of the polydomain sample alone at 0.80 W cm⁻² was 76.1 °C. The last three points of the temperature vs. intensity curve of the polydomain sample with the actuator in front of it were taken and linearly interpolated to find the intensity value that would correspond to a temperature of 76.1 °C, which was 4.73 W cm⁻². NIR Absorbance was calculated as log₁₀(I/I₀), where I = 4.73 W cm⁻² and I₀ = 0.80 W cm⁻².

Calculation of Work Capacity and External/Internal Efficiency

Weights were hung from a small binder clip which was attached to the bottom of the LCE-AuNR actuator. The top of the actuator was secured with a piece of tape. The distance the weight moved was measured using image analysis (see Quantification of Strain by Image Analysis). Work output (J) was calculated as force (N) times distance (m). Work capacity (J kg⁻¹) was calculated as work output divided by the weight of the actuator (1.93 x 10⁻⁵ kg). Input energy (J) was calculated by multiplying the power (W) of the laser and the irradiation time (s). External efficiency was calculated as work output divided by input energy. Internal efficiency was only estimated for the sample carrying a load of 37.9 g and irradiated with 5.10 W cm⁻² NIR light. Under this load, the sample stretched to ~2x the original length. Thus, the absorbance of the stretched actuator was estimated to be half that of the 250 μm thick LCE-AuNR actuator (0.77). The absorbed input energy (J) was calculated by multiplying the Input Energy and NIR Absorbance (0.385). Internal Efficiency was calculated as the Work Output divided by the Absorbed Input Energy.

Quantification of Strain by Image Analysis

Videos taken using the Canon video camera were imported into ImageJ (FIJI), downsampled to 1 frame per second, and saved as .GIF. The plugin MTrackJ was used to track the movement in real time. For work capacity and efficiency calculations, the first point (in the first frame before the NIR laser was turned on) was placed where the actuator and binder clip carrying the load met. In the final frame before the NIR laser was turned off, the point was again placed where the actuator and binder clip carrying the load met. The distance the weight was lifted in pixels was calculated as the difference between the y-axis values of the two points. Distance in pixels was converted to millimeters using the ruler placed next to the actuator as a scale. For actuation in series, a small binder clip weighing ~1 g was attached to the bottom of each actuator to provide some tension to the sample and prevent twisting during actuation. In the first frame, the first point of the three tracks (corresponding to the three actuators) were placed at the top of each actuator. In subsequent frames, points for each track were placed where the actuator and binder clip met. The y-axis values were extracted from each point, and the initial length of each actuator in pixels was calculated as the difference between the y-axis values in the first and second frames. The length in subsequent frames was calculated in the same manner, and the change in length divided by the initial length times 100% was taken as the % strain. For actuation of +1 disclinations, three vertices of the pyramid were tracked (top, bottom left, bottom right). Height was calculated as the difference in the y-coordinates of the bottom left and the top vertices. Length was calculated as the difference in the x-coordinates of the bottom left and bottom right vertices. A custom Python script was then used to calculate the I-unit value according to equation (1) where H is height and L is length. Percent strain was calculated by dividing I-units by 1000.^[35]

$$I={[(\pi ∕2)(H∕L)]}^2\ast {10}^5$$ (1)

Data Analysis and Visualization

GraphPad Prism 9 was used to obtain the derivatives of certain data (% strain s⁻¹, °C s⁻¹, and % strain °C⁻¹). Python and Adobe Illustrator were used to plot the data.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.