Functional stimuli responsive hydrogel devices by self-folding

The development of soft polymeric or hydrogel stimuli responsive actuators and related devices are important from both an intellectual and technological perspective [1–8]. Stimuli responsive soft-actuators are widely observed in nature, for example in the spontaneous and reversible folding and unfolding movements of plants. Examples include the Rhododendron leaves which respond to temperature [9] or the Dionaea muscipula (Venus flytrap) which responds to mechanical and pH stimulation [10, 11]. These actuators enable functions such as defense from predators or consumption of food which are critical to their survival. On the technological front, the development of such bio-inspired synthetic stimuli responsive soft devices could facilitate enhanced functionality for biosensing, robotics and surgery. Specific advantageous traits of these systems include the use of mechanical energy stored within the material, the ability to respond autonomously to enable smart behaviors, and the possibility for extreme miniaturization and untethered operation without the need for wires or batteries or external power sources.

Poly N-isopropylacrylamide (pNIPAM) is an important stimuli responsive hydrogel that undergoes a hydrophilic (swollen in aqueous media) to hydrophobic (deswollen) transition at a temperature just below physiological temperatures (LCST, between 32 °C and 36 °C) [12–16]. Further, when combined with other materials either by co-polymerization, blending or layering, these hydrogels can be made responsive to heat [17–36], pH [37], light [17, 19, 21, 25, 32], mechanical stress [28], ionic strength [37, 38] and magnetic field [35, 39]. Consequently, NIPAM based hydrogels have been used in soft-robotics [18, 24, 32, 35, 37–39], drug delivery [21, 22, 26, 27, 30, 31, 35] and surgery [28].

Despite these advances, the mass-production and miniaturization of such soft cross-linked hydrogel actuators can be challenging. Further, heterogeneous integration with conventional micro-electro-mechanical systems (MEMS) is often required for functional systems which necessitate the development of processes to pattern these soft materials using high resolution and high throughput techniques such as photolithography, imprinting and molding. Precise structuring of complex three-dimensional (3D) structures is also needed for practical applications. An emerging methodology that overcomes these challenges is the self-folding of precisely patterned hydrogel thin films [4, 6, 7, 29–38]. In these methods, existing planar lithographic approaches can be leveraged to create well defined patterns with high resolution and reproducibility. In contrast to previously described relatively higher temperature self-folding approaches for metallic or polymer patterns [40–42], self-folding of hydrogels can be accomplished at physiologically relevant temperatures in aqueous media. 3D actuation with large strains can be enabled by incorporating differential swelling of bilayers or films with compositional or cross-link gradients. For example, Klein et al showed how compositionally heterogeneous pNIPAM films could be used to create buckling sheets [23]. Ionov et al described thermo-responsive microcapsules based on pNIPAM-ABP/PCL bilayers [30, 31]. Kim et al demonstrated buckled gel surfaces using differential swelling of pNIPAM hydrogels using half-tone lithography [33, 34]. Bassik et al showed how pNIPAM-AAc/PEODA bilayers could be used to create flytrap like actuators [37].

Here, we demonstrate a new approach to create pNIPAM based self-folding systems as schematically represented in figure 1. Specifically, we utilized co-polymerized pNIPAM-AAc hydrogels which reduces the response time as compared to single component pNIPAM hydrogels due to increased hydrophilicity [43, 44] and the AAc component also makes the hydrogel pH responsive [45]. Our approach utilizes the fact that bending of hydrogel thin films depend on three important factors namely their thickness, modulus, and swelling ratio. We create two distinct regions within the hydrogel films; flexible hinges and rigid panels. In addition to being thinner, the flexible hinge also has a cross-link gradient [46] while the interconnecting panels are fully cross-linked and thicker. There are several advantages of our approach. As compared to many previous demonstrations of self-folding hydrogel actuators with bilayers, we use a single material that facilitates easy characterization and modelling for the homogeneous material. In addition, use of a single layer eliminates any risk of delamination that can be a concern in bilayer actuators. We demonstrate feasibility of our approach and discuss a finite element model that can be used to predict fold angle. As compared to previous methods such as half-tone lithography [33], the thinner gradient hinge provides increased flexibility and larger fold angles, as illustrated in the model. Finally, we highlight several applications to illustrate practical applicability of the approach.

Zoom In Zoom Out Reset image size

As illustrated briefly in figure 1 and in a more detailed manner in figure S1⁵ , our fabrication approach utilizes two spin coating and two photopatterning steps; (a) low energy UV light exposure on thin hinges and (b) high energy UV light exposure on thick panels. Different types and combinations of patterns can be created by the appropriate design of photomasks for thick panels and hinges.

Details of the process are as follows. We thermally evaporated 20 nm of Cr adhesion promoting and 200 nm Cu sacrificial layers atop commercial silicon wafer substrates. We then spincoated (Specialty Coating Systems, G3P-8 model) the pNIPAM-AAc solution at 600 rpm to deposit the thick panels which were cross-linked by a high energy (500 mJ cm⁻²) UV exposure, through a mask that was appropriately designed using AutoCAD. We exposed the pNIPAM-AAc solution in non-contact mode by separating the mask from the substrate by thin spacers. We utilized a commercial mask aligner (ultra μline series: Quintel) with a 350 W mercury vapor lamp and the exposure energy was measured by multiplying the exposure time with the UV exposure intensity measured using a UV power meter (Vari-Wave II, 365 nm sensor; Quintel). After UV exposure, we developed the thick panels by immersion in acetone for around 7 s, rinsing with IPA and then drying using nitrogen gas. We then patterned the thin hinges using a similar process except that we deposited the film at a higher spin speed to create a thinner film. Also, we utilized a second low energy (50 mJ cm⁻²) UV exposure to photopattern a cross-link gradient along the thickness of the hinges. Then, we developed the hinges as before. We released the patterned panel and hinge structures by dissolving the Cu sacrificial layer in a commercial Cu etchant (APS 100; pH = 2). We observed that the samples could be stored and remained unfolded up to a year in APS 100. Self-folding was trigged only on exposure to the correct thermal or pH conditions.

In order to accurately target fold angles, we first measured post-cross-link thickness of rectangular test structures versus spin speed curves at the same UV exposure (500 mJ cm⁻²). We dispensed the pNIPAM-AAc hydrogel solution at six different spin speeds ranging from 600 rpm to 1200 rpm onto flat Si wafer substrates. We used a color 3D pinhole confocal optical microscope (KEYENCE VK-X100K) to image and measure the thicknesses of photopatterned pNIPAM-AAc test structures. Our thickness versus spin speed data is plotted in figure 2(A) and our data demonstrates that the thickness could be reproducibly controlled between thin 1.6 μm panels at the high spin speeds of 1200 rpm and thick 6.5 μm panels at low spin speeds of 600 rpm. We note that these films are thinner than the previously utilized pNIPAM self-folding structures which typically had thickness between 10 μm [33] and 225 μm [37]. We also measured the resolution of our photopatterning process using a ruler mask and the smallest feature that could be reproducibly patterned was 8 μm (figures 2(B), (C)). Color images in figures 2(D)–(F) show that different hinge and panel thickness could be implemented and also that the thickness within the fully cross-linked panels was relatively uniform.

Zoom In Zoom Out Reset image size

By varying the patterns of thick fully cross-linked panels and thin gradient cross-linked hinges using AutoCAD designed photomasks, we were able to create a variety of structures. After fabrication and release from the substrate, the planar photopatterned structures were heated up to 60 °C in the copper etchant APS 100 solution (pH 2). Illustrative examples of a self-folded pyramidal and cubic shaped capsule are shown in figure 3. The second and third columns in figures 3(A), (B) shows the intermediate and final self-folding steps. We observed that the yield was approximately 60% for well folded structures with no discernible defects when visualized by optical microscopy.

Zoom In Zoom Out Reset image size

A key element of our approach is the creation of cross-link gradients on low UV exposure of 50 mJ cm⁻². The influence of such low and high exposure energy is evident in the patterning of hole sheets shown in figures 4(A) and (B). Such sheets of identical thickness but exposed to low and high energy showed different self-folding characteristics. While sheets exposed to 500 mJ cm⁻² buckle, they do not roll up due to the absence of a cross-link gradient (figure 4(C)). In contrast, identically shaped sheets patterned with a lower energy of 50 mJ cm⁻² roll up due to a cross-link gradient along the thickness of the film (figure 4(D)). We characterized the cross-link gradients by adding a fluorescent dye (rhodamine 6 G) into the pNIPAM-AAc solution which was then sonicated for 30 min to form a well-mixed solution. The gradients in the films patterned with low UV exposure could then be visualized using confocal fluorescence microscopy (LSM 510 META, Zeiss). Specifically, we measured the relative pixel intensities from the top to bottom confocal images by depth discriminations and an intensity gradient was observed (figure 4(E)). This experimental data validates our approach of creating hinges with cross-link gradients for self-folding.

Zoom In Zoom Out Reset image size

In addition to experiments, we investigated the thermo-responsive deformation mechanism of the pNIPAM-AAc hydrogel via a 3D constitutive finite deformation model. The model assumes that the free energy change of the hydrogel system is caused by the mechanical stretching of polymer chains based on the classical Gaussian statistics and the polymer and solvent mixing based on Flory-Huggins model. Parameters of this model included the shear modulus and temperature-dependent Flory–Huggins interaction parameter. To determine the interaction parameter, the swelling mass ratio of cross-linked hydrogels was experimentally measured (figure 5). These measurements were carried out by spreading 500 μl pNIPAM-AAc solutions on a glass substrate, exposing the films to UV light with the relevant energies of 50 and 500 mJ cm⁻² and soaking in DI water. We ensured that the samples reached equilibrium swelling by placing them on a hot plate at each temperature for at least 12 h.

Zoom In Zoom Out Reset image size

The weight of the gel at this stage represents that of the swollen gel. The swelling ratio $\Omega =\frac{Wt.\;of\;swollen\;gel}{Wt.\;of\;deswollen\;gel}$ was calculated by dividing the measured weight of the swollen gel at each temperature to that of the deswollen gel (dehydrated gel).

In addition to swelling ratios, the shear modulus of the swollen pNIPAM-AAc hydrogels was measured using a dynamic mechanical analyzer (Q800 DMA; TA instruments). The fully swollen samples exposed at high and low energies were both measured at room temperature in the dynamic mode, with a frequency sweep from 1 Hz to 10 Hz, and a 0.5% applied strain. The storage modulus was independent of the frequency and is equivalent to the equilibrium Young's modulus. The shear modulus G is related to the Young’s modulus E as: $G=\frac{E}{3}{\mkern 1mu} \Omega _{0}^{1/3}$ where Ω₀ is the swelling ratio at room temperature.

The constitutive model for the mechanical and swelling behavior of hydrogels is described below. Consider the deformation gradient, defined as F = ∂ x/∂ X, which describes the stretch and rotation of material lines from the dry stress-free body to the swollen and stressed body. The deformation gradient is decomposed into a mechanical part F_e and an isotropic swelling part F_s as $\;{\boldsymbol{F}} ={{\varphi }^{-{\bf 1}/{\bf 3}}}{{{\bf F}}_{{\boldsymbol{e}} }}$, where φ is the polymer volume fraction of the polymer-solute system at the swollen, stressed state. Since the gel is initially swollen, we define the initial swollen state as the stress-free reference state, and a deformation gradient f mapping the points from the stress-free reference state to the final spatial points:

$$\begin{eqnarray}&&{\bf f}=\varphi _{0}^{{\bf 1}/{\bf 3}}{\bf F}\end{eqnarray} \tag{ 1 }$$

where φ₀ is the polymer volume fraction of the polymer-solute system at the reference state. The polymer volume fraction is represented as φ = 1/ (1 + vc) [47], where v is the volume per solute molecule and c is the number of solute molecules per reference volume. We define the left Cauchy-Green tensor as b = FF^T and express it in terms of its eigenvalues (principal stretches) and eigenvectors (principal directions):

$$\begin{eqnarray}&&{\boldsymbol{b}} =\sum \limits_{{\boldsymbol{a}} =1}^{3} {\boldsymbol{ \lambda }} _{{\boldsymbol{a}} }^{2}\;{{{\boldsymbol{n}} }_{{\boldsymbol{a}} }}\otimes {{{\boldsymbol{n}} }_{{\boldsymbol{a}} }}\end{eqnarray} \tag{ 2 }$$

where ${{\lambda }_{a}}=\varphi _{0}^{-1/3}{{\bar{\lambda }}_{a}}$ and ${{\bar{\lambda }}_{a}}$ are the corresponding principal stretches of f. We assume that free energy density of the total system is caused by the stretching of polymer chains and the mixing of polymer and solvent, and is represented as:

$$\begin{eqnarray}&&\Psi ={{\Psi }_{e}}\left( {\bf F},\varphi \; \right)+{{\Psi }_{m}}(\varphi )\end{eqnarray} \tag{ 3 }$$

We adopt a quasi-incompressible model for the free energy density associated with stretching polymer chains [48]

$$\begin{eqnarray}&&\begin{array}{lllllllllllllll} {{\Psi }_{e}}=\frac{G}{2}\left\{ \lambda _{1}^{2}+\lambda _{2}^{2}+\lambda _{3}^{2}-3-2{\rm ln} \left( {{\lambda }_{1}}{{\lambda }_{2}}{{\lambda }_{3}} \right) \right\} \\ \quad \quad +\frac{\kappa }{4}\left\{ {{\left( \varphi {{\lambda }_{1}}{{\lambda }_{2}}{{\lambda }_{3}} \right)}^{2}}-2ln\left( \varphi {{\lambda }_{1}}{{\lambda }_{2}}{{\lambda }_{3}} \right)-1 \right\} \\ \end{array}\end{eqnarray} \tag{ 4 }$$

where G and κ are the shear modulus and bulk modulus respectively. We assumed the bulk modulus is 1000 times of the shear modulus to achieve the volumetric incompressibility of mechanical deformation.

The mixing component of the free-energy density is represented as [49, 50]

$$\begin{eqnarray}&&\;{{\Psi }_{m}}=\frac{RT}{\nu \varphi }\left\{ \left( 1-\varphi \right)ln\left( 1-\varphi \right)+\chi \varphi \left( 1-\varphi \right) \right\},\end{eqnarray} \tag{ 5 }$$

where R is the gas constant and χ is the Flory–Huggins interaction parameter. The Cauchy stresses tensor σ and the chemical potential μ can be obtained from the free energy [47, 51], as

$$\begin{eqnarray}&&{\bf \sigma }=\sum \limits_{a=1}^{3} \frac{1}{{{\lambda }_{1}}{{\lambda }_{2}}{{\lambda }_{3}}}{{\lambda }_{a}}\frac{\partial \Psi }{\partial {{\lambda }_{a}}}\;{{{\boldsymbol{n}} }_{a}}\otimes {{{\boldsymbol{n}} }_{a}}\end{eqnarray} \tag{ 6 }$$

$$\begin{eqnarray}&&\mu =\frac{\partial \Psi }{\partial {\bf c}}\;\end{eqnarray} \tag{ 7 }$$

Substituting equations (3)–(5) into equations (6) and (7) yields,

$$\begin{eqnarray}&&{\bf \sigma }=\sum \limits_{a=1}^{3} \left\{ \begin{array}{lllllllllllllll} \frac{G{{\varphi }_{0}}}{\;{{{\bar{\lambda }}}_{1}}{{{\bar{\lambda }}}_{2}}{{{\bar{\lambda }}}_{3}}}\left( \varphi _{0}^{-2/3}\bar{\lambda }_{a}^{2}-1 \right) \\ \quad +\frac{\kappa {{\varphi }_{0}}}{2\;{{{\bar{\lambda }}}_{1}}{{{\bar{\lambda }}}_{2}}{{{\bar{\lambda }}}_{3}}}\left[ {{\left( \varphi \varphi _{0}^{-1}{{{\bar{\lambda }}}_{1}}{{{\bar{\lambda }}}_{2}}{{{\bar{\lambda }}}_{3}} \right)}^{2}}-1 \right] \\ \end{array} \right\}{{{\boldsymbol{n}} }_{a}}\otimes {{{\boldsymbol{n}} }_{a}}\end{eqnarray} \tag{ 8 }$$

$$\begin{eqnarray}&&\begin{array}{lllllllllllllll} \mu =RT\left\{ ln\left( 1-\varphi \right)+\varphi +\chi {{\varphi }^{2}} \right\} \\ \quad \quad -\frac{\kappa \nu \varphi }{2}\left\{ {{\left( \varphi \varphi _{0}^{-1}\;{{{\bar{\lambda }}}_{1}}{{{\bar{\lambda }}}_{2}}{{{\bar{\lambda }}}_{3}} \right)}^{2}}-1 \right\} \\ \end{array}\end{eqnarray} \tag{ 9 }$$

To model the thermo-responsive effect, we assumed that the temperature dependent Flory–Huggins interaction parameter χ has the following form [51],

$$\begin{eqnarray}&&\begin{array}{lllllllllllllll} \chi =\frac{1}{2}\left( {{\chi }_{{\rm L}}}+{{\chi }_{{\rm H}}} \right) \\ \quad \quad +\frac{1}{2}\left( {{\chi }_{{\rm H}}}-{{\chi }_{{\rm L}}} \right){\rm tanh} \left( \frac{T-{{T}_{{\rm tran}}}}{\Delta T} \right) \\ \end{array}\end{eqnarray} \tag{ 10 }$$

where χ_L and χ_H are the Flory–Huggins interaction parameters at low temperature and high temperature respectively, T_tran is the transition temperature and ΔT is the width of the transition region. These parameters were obtained by fitting the simulation free swelling data to the experimental results as shown in figure 5. The parameters used in the model are listed in table 1.

The constitutive model was implemented into TAHOE (Sandia National Laboratories) to simulate the self-folding pyramidal and cubic structures. Figure 6 shows the finite element model of the pyramidal specimen (the finite element model of cubic specimen is not shown). The mesh was discretized using trilinear hexahedral elements and a higher mesh density was chosen at the hinge. For simplicity, we modeled the gradient in the hinges as a two-layer structure where the top layer of the hydrogel has a higher cross-linker density and corresponding smaller swelling ratio (G = 90 KPa and χ_L = 0.58) compared with the bottom layer with a lower modulus and higher swelling ratio (G = 60 KPa and χ_L  =  0.562). The parameters for the model were chosen to reproduce the modulus and swelling ratio measured for the 50 mJ cm⁻² materials shown in table 1. The displacement boundary conditions were set as:

$$\begin{eqnarray*}&&{{u}_{x}}\;\left( x=0,y,z \right)=0,{{u}_{y}}\;\left( x=0,y=0,z \right)=0,\end{eqnarray*}$$

$$\begin{eqnarray}&&{{u}_{z}}\;\left( x=0,y=0,z=0 \right)=0.\end{eqnarray} \tag{ 11 }$$

Zoom In Zoom Out Reset image size

The simulation starts from stress-free state at temperature T = 25 °C, at which temperature the structure is unfolded as shown in experiments. The initial polymer fraction φ₀ is obtained by solving equations (8) and (9) with the condition σ = 0 and μ = 0. The temperature is continuously increased to 37 °C. At each temperature, the deformation gradient f and polymer concentration φ are updated to satisfy the force balance and chemical potential equilibrium. From the simulation, we can obtain the folding angle as a function of temperature (figure 7(A); video S1A and B in supplementary⁵). To verify the model, we measured the folding angles of cubic structures. We did not measure the folding angle of the pyramidal structure because it was difficult to obtain an accurate angle from experiments. As shown in figure 7(A), the simulation shows good agreement with the experimental data. We also did a parameter study by changing the thickness of hinge. As expected, a thinner hinge yields a large folding angle as shown in figure 7(B). In the present study, the thinnest hinge we could reproducibly deposit by spin coating was 1.6 μm which was used. If alternate techniques such as layer by layer (LBL), surface initiated polymerization or self-assembly are used, then even thinner hinges could possibly be fabricated for even greater fold angles.

Zoom In Zoom Out Reset image size

Since, we can create a wide range of structures using this approach we highlight a few applications. In figure 8, we show self-folding of a cubic capsule that assembled on heating. Figures 8(A), (B) indicates the high-throughput nature of the process so that many structures can be patterned and assembled at once. Once assembled, the structures can be loaded with cargo such as the beads shown in figure 8(C). This example highlights possible utility of this approach to create self-folding capsules or microcontainers. Further, transparency of the pNIPAM-AAc faces permits visualization of the encapsulated cargo using optical bright field or fluorescence microscopy.

Zoom In Zoom Out Reset image size

In addition to capsules, we also highlight the combined thermally and pH responsive opening and closing of grippers (figure 9). The grippers were patterned with six symmetrical fingers and three joints in each finger. The grippers close in environments of pH 2, T = 60 °C and open in environments of pH 7, T = 25 °C. The reversibility of these pNIPAM-AAc grippers was tested over several folding to unfolding cycles and no phenomenological residual plastic deformation was observed under parallel stimuli triggers between (60 °C, pH 2) and (25 °C, pH 7) (supporting video S2⁵). We did however observe that the soft digits sometimes adhered to each other or the underlying substrate.

Zoom In Zoom Out Reset image size

Finally, stimuli responsive micro-mirror scanners were designed to demonstrate proof-of-concept applicability in optics (figure 10). Recently, much research has been directed at the design and fabrication of smart material based optomechanical micro-mirrors [52]. Additionally, micro-mirror elements have been widely used in optical systems such as flat panel displays, optical interconnects, adaptive optical arrays and laser beam scanners [53, 54]. They benefit from low weight, low operating optical power, compact design and sizes, which are potentially applicable to a variety of optoelectronic devices. We created five-faced pNIPAM-AAc structures with square (500 μm × 500 μm) Au mirrors that were attached using an epoxy. In order to highlight applicability of a thermally responsive micro-mirror steering system, we directed a beam of light at these structures. As depicted in figure 10(A), a 650 nm red laser beam was expanded and collimated by a beam-expander and the collimated beam was coupled into 50 μm diameter multimode fiber core by a microscope objective lens (Plan 10x/0.25, Olympus). The distal end of the fiber was directed toward the gold mirror mounted on the hydrogel actuator (figure 10(B)). The light reflected by the micro-mirror through the 200 μm multimode fiber was detected using an avalanche photodetector (APD110A, Thorlabs). When the structures self-fold in different environments, the position of the beam of light changes from reflection from a single mirror to two mirrors separated by different distances as can be seen in figures 10(C), (D). In the present design, the distance between incident and reflected beam vary from d = 0 mm to 2 mm. Hence, both the intensity and direction of the reflected beam varies with the environment (temperature, pH) and can be utilized for steering in optical beams or to decipher positional information. The advantages are that this approach is contactless and does not require wired connections or external power sources.

Zoom In Zoom Out Reset image size

In conclusion, we have developed a new approach to enable self-folding stimuli responsive microstructures. Photolithographic patterning of thin gradient cross-linked hinges and thick fully cross-linked panels have enabled advanced functionalities such as capsules, grippers and stimuli responsive micro-mirrors. We anticipate widespread applicability in optics, electronics, and robotics. A predictive finite element model was used to rationalize fold angles and its variation with hinge thickness and swelling ratio. Our study also highlights several important challenges. Notably increasing the speed of actuation to the sub-second time scales remains challenging. Additionally, increasing the resolution of lithographic patterning such as by imprint techniques could enable smaller structures. Finally due to the low modulus of hydrogels, such actuators are floppy and additionally work primarily in aqueous environments. The inclusion of more rigid elements [36] and microfluidic interfaces could be used to further expand their capabilities.

4.1. Materials

4.2. Synthesis of pNIPAM-AAc

4.3. Imaging of pNIPAM-AAc hydrogel structures

This work was supported by the National Science Foundation grant NSF CBET-1066898 (to DHG).