Amongst all stimuli-responsive shape-shifting polymers, shape memory polymer (SMP)8-13 is uniquely attractive since it can be programmed to transform from an arbitrary temporary shape to its permanent shape. Typically, the shape transformation starts as soon as a stimulus is exerted14. Although this on-demand response is beneficial in many circumstances, the need to apply external stimulation has increasingly become the bottleneck for many demanding applications. For instance, SMP based medical devices have over the years become attractive options for minimum invasive surgery8. There, access to active heating or other forms of stimulation is a major hurdle. Using body heat as the natural trigger is a potential option, but is limited by the uncontrolled deployment during the delivery path that can lead to device trapping at an unintended location. This issue is schematically illustrated in Fig. 1a, showing that its shape transformation starts quickly after the material reaches the target temperature. The onset of transformation as dominated by thermal conduction is too short (typically seconds) for device control.
Non-heating triggers for water-induced15 or degradation-induced16 SMP are other options. However, water diffusion and degradation are exceedingly slow and geometry dependent, making it difficult to control at the device level. Temporal programming11,12 offers a drastically different alternative. It utilizes supramolecular interactions as the molecular switches for the shape transformation. Without requiring any physical triggers, its shape-shifting can occur autonomously with its kinetics controllable by tuning the time and temperature during the programming. However, temporally programmable SMP typically show undesired large instantaneous shape changing at the beginning of the transformation process (Fig. 1a).
The externally triggered on-demand SMP17-19 and the temporally programmable autonomous SMP11,12 both have their own advantages and limitations. An ideal SMP should possess their respective desired attributes in one material, that is, autonomous shape transformation in an on-demand fashion. At first glance, achieving such a behavior is infeasible given that these two characteristics are intrinsically conflicting. However, if a molecular mechanism can be devised to have on-demand control of the onset of autonomous shape-shifting, this conflict can be potentially resolved. For such an SMP (Fig. 1a), not only can it transform autonomously, but that its onset can be controlled in a programmable way to offer the on-demand manipulation.
To achieve this goal, the enabling shape memory mechanism should differ from the thermal phase transition (e.g. glass transition and melting transition) behind conventional SMP 17-19 such that the shape-shifting onset is no longer determined by thermal conduction. Spinodal decomposition comes to our mind as it allows phase separation to occur spontaneously with the kinetics dominated by internal mass diffusion instead of thermal conduction. This principle is potentially applicable to many multi-component polymers. We pick a hydrogel with heating-induced spinodal decomposition20 to test our hypothesis. Our other goal is to introduce 3D printability as this would provide the freedom to access complexed permanent geometries key to SMP applications. For this reason, we design an aqueous photo-curable resin (Fig. 1b) consisting of acrylic acid (AAc), N,N’-methylenebisacrylamide (BIS), photo-initiator (I819), and water. After photo-curing (Fig. 1c), the material is soaked in calcium acetate (CaAc) solution and then deionized water to obtain the target hydrogel. A series of samples with different BIS contents were synthesized and their mechanical properties are presented in Supplementary Fig. 1. Considering its optimal mechanical properties, the sample containing 0.25 mol% of BIS crosslinker was chosen for further investigation. Its modulus is 1.12 GPa at 90 °C, which drops to 314 kPa at 25 °C (see mechanical curves in Supplementary Fig. 2). This significant softening under natural ambient temperature (25 °C) implies its potential for autonomous shape transformation. Importantly, the hydrogel volume remains nearly unchanged during the heating/cooling process (Supplementary Fig. 3), suggesting internal molecular event(s) for the rigidity change.
To probe the shape memory property (Fig. 2a), a flat hydrogel sheet is mechanically deformed into a “windmill” shape at room temperature. Heating at 90 °C while maintaining the deformation force allows fixing this temporary shape. Subsequent removal of the force completes the programming process. When transferred back to 25 °C, the windmill does not recover immediately, it instead starts 10 min later (Supplementary Video 1). This is rather notable since temperature monitoring using an infrared camera shows that the sample reaches the equilibrium temperature within 10 seconds (Fig. 2b). In essence, the shape recovery process is not governed by heat conduction. Instead, it is dominated by the spinodal decomposition process (Fig. 2c). At ambient temperature, the polymer chain mobility in the hydrogel is high, resembling the rubbery state of conventional SMP. Upon heating, the hydrogel undergoes spinodal decomposition. The chains are packed into condensed polymer domains with limited chain mobility, analogous to the glassy state of conventional amorphous SMP. Critically, the “glassy to rubbery” transition for the hydrogel is dominated by mass diffusion instead of heat conduction. When the hydrogel is cooled to ambient temperature, the polymer chains do not regain their mobility immediately since the redistribution of internal water is mass transport limited. Correspondingly, the macroscopic shape recovery can only start when the mass diffusion proceeds to a certain level.
To verify the above mechanistic hypothesis, the morphology change of the hydrogel is monitored (Fig. 2d). The hydrogel displays an original homogenous morphology at 25 °C. Heating at 90 °C for 180 min results in strong phase separation. Upon natural cooling for 10 min, the feature size of the separated phase regions reduces from 50 μm to 10 μm, corresponding to the beginning of the macroscopic shape recovery. Eventually, the morphology reverts back to the homogeneous state and the shape fully recovers. The morphological change is accompanied by modulus change (Fig. 2e). The modulus declines continuously upon cooling and reaches 141 MPa after 10 min. At this state, the chains are still stiff enough to prevent the shape recovery. With further water redistribution, the modulus decreases further and the polymer chains become sufficiently activated for shape recovery.
The next question is: can the onset period be modulated? Here, the heating-induced phase separation is exactly opposite to the vanishing process upon natural cooling. We therefore hypothesize that the degree of spinodal decomposition can be modulated by heating time to potentially allow programming the onset of the autonomous shape transformation. Indeed, Fig. 3a verifies that the phase separation upon heating gets stronger with time, evidenced by the progressive increase of the feature size of the separated regions. We further speculate that, for stronger phase separation achieved during the shape programming step, water redistribution across the phase regions in the autonomous recovery step would require more time. This should lead to a longer onset period for the shape recovery. This is experimentally proven to be valid (Fig. 3b). The recovery onset indeed increases notably for a longer heating time during programming. Quantitatively, we define the onset period (tonset) as the time corresponding to 5% recovery. Accordingly, tonset can be finely tuned from 1 min to 22 min when the heating time varies from 1 min to 400 min (Fig. 3c). Heating beyond 400 min does not increase tonset further, indicating that the system reaches the equilibrium phase separation state. In addition, the recovery speed gets lower when the heating time increases, but full recovery can be achieved in all cases (Supplementary Fig. 4).
The crosslinking density of the hydrogel is also a key parameter to tune tonset. Increasing crosslinking density results in decrease in enthalpy (Fig. 3d and Supplementary Table 1), implying that the phase separation gets increasingly more difficult. However, tonset does not change monotonically with the crosslinking density (Fig. 3e). Here, higher crosslinking leads to weakened phase separation which favors faster water redistribution, but it also results in reduced chain mobility which impedes water redistribution. The trade-off is such that the three samples with 0.25-2.5 mol% crosslinker display the shortest tonset whereas those with a higher or lower crosslinker have much longer tonset (Supplementary Fig. 5). Another parameter, the programming strain, exhibits does not impact tonset (Supplementary Fig. 6). This is desirable for geometrically complexed devices for which the strain is typically non-uniform.
The above mechanism centers on internal water redistribution. To preclude the impact of surrounding water, identical experiments are performed using oil as the heating/cooling medium. The change does not notably alter the tonset (Supplementary Fig. 7). tonset is also largely independent of the sample thickness (Supplementary Fig. 8). The above two features differ notably from common shape memory hydrogels21-24 for which the shape changing kinetics is strongly affected by the device dimension due to the inside-and-out water diffusion25. This reliance on internal water redistribution for the current system offers much better control for geometrically complex devices produced by light curing based 3D printing, with the autonomous shape memory behavior well preserved (Fig. 4a). Closer monitoring of the shape recovery process (Fig. 4b) clearly shows uniform shape-shifting onset around 10 minutes (Fig. 1c) despite the varied thickness in different parts of the device.
The combination of 3D printing and the on-demand autonomous shape-shifting offers unique advantages beyond other 4D printing systems26-28. Fig. 4c illustrates that it provides a much-needed time window for device delivery into a complexed vessel immersed in deep water which mimics human body environment. For a classical SMP stent (Tg: 40 °C)26 triggered by heating from the surrounding water (70 °C), the implantation fails due to the premature deployment during the delivery path (Fig. 4c and Supplementary Video 2). In contrast, the time-delayed device recovery for the current system allows successful deployment in the intended location at 25 °C (without heating). Fig. 4d and Supplementary Video 3 illustrate another benefit for potential bio-probe applications. Here, the probe needs to be rigid to puncture through the tough barriers (e.g. surface tissues) and soft enough to match the low modulus of the target tissue. The soft gelatin core embedded in much stiffer silicone rubber mimics such a situation. Without programming, the hydrogel device is too soft to penetrate through the silicone. After programming at 90 °C, the device becomes rigid for sufficiently long time at 25 °C. This allows successful penetration while deploying at a specific time when it is fully in the soft gelatin. Fig. 4e demonstrates that tonset can be programmed spatially to realize unique sequential deployment (Supplementary Video 4). The device is programmed uniformly at 90 °C for 1 hour. The three regions (i, ii, and iii) are then immersed in 25 °C water for 6, 3, and 0 minutes, respectively. When the entire device is finally immersed in 25 °C water, the three regions recover in a sequential order. This autonomous multi-shape transformation is drastically different from existing examples that would otherwise require complexed stimulation schemes (e.g. multi-step heating17,18) that are difficult to implement for many applications.
The above versatile shape-shifting behaviors are achieved in the easily controllable manner. The underlying spinodal decomposition mechanism, while generally known in polymer science, represents a new molecular tool for designing SMP. It allows decoupling shape transformation from the heat transport dominated process for common SMP, consequently resolving the conflict between on-demand control and autonomous shape-shifting. The internal mass diffusion mechanism also differs from other shape memory hydrogels, offering another degree of freedom. The general principle described here can potentially inspire future design of other types of shape-shifting materials.