Adaptive memory of hydrogels with tunable hysteresis

Abstract

The creation of adaptive memory based on soft matter, similar to the brain, is an attractive and challenging research area. Hysteresis is closely related to adaptive memory because it involves a system's ability to retain and utilize information about its past states or inputs to influence its current and future behavior. To achieve adaptive memory control, it is highly desirable to develop stimuli-responsive hydrogels with a tunable hysteresis in the volume phase transition. Herein, we propose a one-pot synthesis method to develop environmentally adaptive memory by preparing dual-responsive hydrogels (e.g., poly(N-isopropylacrylamide-co-acrylic acid)-g-methylcellulose). The range of the hysteresis window in temperature-dependent shape morphing can be adjusted from approximately 0 °C to 17.6 °C by changing the pH stimulus. Furthermore, the thermal hysteresis windows adapt to the surrounding temperature autonomously. The P(NIPAm-co-AAc)-g-MC hydrogel can maintain a series of small hysteresis loops, which are suitable for memorizing multiple states. Applications in microvalves, hydrogel patterns and smart windows are successfully demonstrated, leveraging the intrinsic hysteresis behavior of the hydrogels. The memory function can switch between memorizing and forgetting behavior, and the memory window adapts to environmental stimuli autonomously. This work contributes an innovative strategy to the development of adaptive memory based on soft materials, paving the way for more intelligent systems.

---

New concepts

The creation of adaptive memory based on soft matter, similar to the brain, is an attractive and challenging research area. Hysteresis can involve a system's ability to retain its past states to influence its current and future behavior. However, most reported hysteresis of stimuli-responsive materials fails to exhibit switching between states with hysteresis and without hysteresis after fabrication, which means the memory function is unable to switch between memorizing and forgetting behavior. Moreover, the memory window of hysteresis is unable to adapt to environmental stimuli (e.g., pH and temperature) autonomously or be adaptively adjusted. Herein, we propose a one-pot synthesis method to develop environmentally adaptive memory by preparing dual-responsive hydrogels (e.g., poly(N-isopropylacrylamide-co-acrylic acid)-g-methylcellulose). The hysteresis window in temperature-dependent shape morphing can be adjusted from approximately 0 °C to 17.6 °C by changing the pH stimulus, which shows that the memory function can switch between memorizing and forgetting behavior. Furthermore, the thermal hysteresis windows adapt to the surrounding temperature autonomously. The P(NIPAm-co-AAc)-g-MC hydrogel can maintain a series of small hysteresis loops, which are suitable for memorizing multiple states. This work contributes an innovative strategy to the development of adaptive memory based on soft materials, paving the way for more intelligent systems.

Introduction

Intelligence can be regarded as the ability to acquire information, comprehend information and retain it as knowledge, which can be utilized in adaptive behavior within a changing environment. The capacity for adapting to surroundings is one of the most important main traits in intelligence and is commonly found in living organisms.¹ An adaptive system is able to dynamically regulate its system parameters instead of maintaining a fixed input to output relationship.² Achieving the capability of memory evolution and adaptability to changing environments is one of the ultimate goals of artificial intelligence.³ The memory of our brain, stored in soft matter, is adaptive, and it remembers or forgets different information. By comparison, existing manmade memory, formed from hard materials, is static, and it does not adjust its memorizing behavior under environmental external stimulation.⁴ Creating adaptive memory based on soft materials, similar to the brain, is an emerging and attractive research area. By receiving and responding to external stimuli, such smart materials would interact with the surroundings, while internally adapting their structure to realize the distribution and storage of information.⁴ Thus, it is very important to develop strategies to achieve adaptive memory, which exhibits the intrinsic ability to process internal feedback and not only modifies information storage, but even regulates memory function in response to changing environments.

In recent years, increasingly advanced soft matter with memory functions has been the focal point in the pursuit of new material applications. Shape memory and stimuli-responsive materials have attracted major efforts in the development of material properties that can be changed on demand by imposed stimuli, such as pH, temperature, light, magnetic field, electrical and mechanical stimuli.^5,6 Shape memory polymers (SMPs) exhibit a permanent shape at low temperature (below the transition temperature), deform at a high temperature (above the transition temperature), and maintain the deformed shape after cooling down (below the transition temperature). SMPs turn back from the deformed shape to the permanent shape upon reheating. Recently, researchers utilized a single medical-grade SMP to combine shape morphing and SMP behaviors at physiological temperatures to yield smart implants with precise control over dimensions for tissue repair and regeneration.⁷ Such a shape memory effect has been found in a variety of polymers, such as liquid crystalline elastomers, amorphous polymers and semicrystalline polymers.⁵ The switching temperature region of SMPs can determine the service temperature of the materials, which is a crucial parameter for the application of SMPs. From the molecular viewpoint, the switching temperature region is generally related to the melting temperature of semicrystalline SMPs or the glass transition temperature of amorphous SMPs.^7–9 The switching temperature region of SMPs can be effectively tuned by varying the photo-crosslinking degrees of the polymer network,¹⁰ copolymer compositions⁹ and material compositions.^11–13 However, these strategies for memory function do not enable switching between memorizing and forgetting behavior, and the switching temperature region fails to be adaptively and reversibly adjusted after fabrication.

A wide range of stimuli-responsive materials that respond to many different types of stimulus has been fabricated and has found promising applications in grippers, sensors, valves, soft robotics and biomedical applications.¹⁴ Recently, multifunctional conductive hydrogels comprising alginate, MC and acid-functionalized carbon nanotubes fabricated by 4D printing were demonstrated as flexible strain sensors. Furthermore, the synergistic effect of electrical cues and the in situ self-folding ability of these hydrogels enabled sutureless neurorrhaphy and superior neural regeneration.¹⁵ Stimuli-responsive materials typically exhibit path-independent, non-volatile behavior, with a one-to-one correlation between the environmental stimulus and their material properties. For example, the material reliably demonstrates a prescribed property at a given temperature, regardless of its prior thermal history. However, these materials fail to achieve memory functionality due to their history-independence. In intelligent systems, memory enables the system to learn from past stimuli and adjust responses accordingly. Hysteresis-like behavior can provide a mechanism for retaining and weighting past inputs, allowing for adaptable responses. Depending on the environmental history of the material, multiple states of the material can exist under the same environmental conditions.^16–18 Hysteresis is an interesting phenomenon, which displays an output depending on both past and present inputs. In addition to ferroelectric and ferromagnetic materials, some soft matter materials such as stimuli-responsive soft materials also exhibit hysteresis.^19,20 Such hysteretic properties of soft materials could be used in memory devices. Bistable states of the soft material can exist, resulting in storage of information on physical history in a single environment.^21,22 Polymers with large thermal hysteresis can memorize historic information and the memory may be removed by excessive cooling, such as a large thermal hysteresis in volume phase transition, which paves the way for the next generation of microvalves, actuators, drug delivery systems and microrobots.^19,23–25 In general, the thermal hysteresis of the LCST, i.e. the difference in the volume phase transition temperatures of the heating and cooling processes, is very small due to the rapid dissolution of the polymer chains in their shrunken state.²⁶ It has been reported that polymers with large thermal hysteresis can memorize changes in their state and that the memory may be deleted by excessive cooling.^27,28 Recently, studies have reported several methods to enhance the thermal hysteresis of materials, which include the incorporation of MC^19,29,30 and amphiphilic groups,³¹ control of stereo-regularity,³² utilization of host–guest interactions³³ and changing of the electrostatic crosslinker.³⁴ These studies mainly concerned sol–gel transitions^25,29 or transmittance changes.^16,29,30,34 To achieve programmable volume control, it is highly desirable to develop stimuli-responsive hydrogels with a large hysteresis capable of memory in the volume phase transition. Mori et al. have previously reported a hydrogel prepared with an α,α-disubstituted vinyl monomer with amphiphilic substituents, which exhibited a 5 °C thermal hysteresis in its volume change.²⁴ This improved hysteresis was possible because the material was capable of forming a metastable aggregate due to its enhanced intra/intermolecular interaction. Dowan et al. reported the preparation of a stimuli-responsive hydrogel by incorporation of MC, which showed a 10 °C thermal hysteresis in its volume phase transition.²² While they demonstrated the capability of influencing a relatively large thermal hysteresis, the materials failed to be switched between exhibiting and not exhibiting hysteresis once fabricated,^19,21,22,34 which means the memory function is unable to switch between memorizing and forgetting behavior. Moreover, most of the previous studies showed that the memory window of hysteresis is unable to adapt to environmental stimuli (e.g. pH and temperature) autonomously or be adaptively adjusted after fabrication. To achieve adaptive memory control, it is highly desirable to develop stimuli-responsive hydrogels with a tunable hysteresis capable of adaptive memory in the volume phase transition. Unlike classic synthetic SMPs and stimulus-responsive materials, which are limited to fixed responses, living organisms exhibit dynamically adaptive responses influenced by the intensity, repetition, and history of stimuli. Therefore, adaptive memory that advances “material intelligence” and progressively life-like properties is an emerging need, and yet a grand challenge.

In this study, we propose a one-pot synthesis to fabricate dual-responsive hydrogels with adaptive memory that exhibits environmental adaptive memorizing behavior. The dual-responsive hydrogels poly(N-isopropylacrylamide-co-acrylic acid)-g-methylcellulose (P(NIPAm-co-AAc)-g-MC) and poly(N-n-propylacrylamide-co-acrylic acid)-g-methylcellulose (P(NNPAm-co-AAc)-g-MC), with tunable thermal hysteresis, were prepared by introducing P(NIPAm-co-AAc) or P(NNPAm-co-AAc) chains onto MC backbones and concurrently crosslinking the P(NIPAm-co-AAc) or P(NNPAm-co-AAc) (Fig. 1(a)). The adaptive memorizing behavior is achieved based on the MC backbone and dual responsiveness. It was anticipated that stabilized aggregates would be formed above the lower critical solution temperature (LCST) due to the intermolecular interactions of the MC backbone. Such stabilized aggregates are very important for achieving a large thermal hysteresis in volume change.¹⁷ The memory window of hysteresis can be tuned by varying the LCST of the P(NIPAm-co-AAc)-g-MC or P(NNPAm-co-AAc)-g-MC hydrogels upon environmental pH stimulation. The memory function is able to switch between memorizing and forgetting behavior, and the memory window can be adaptively and reversibly adjusted after hydrogel fabrication (Fig. 1(b)). Moreover, the different partially contracted states of the hydrogels are also maintained, resulting in a series of relatively small thermal hysteresis loops, which are suitable for adaptively memorizing more states. This strategy of developing adaptive memory based on a tunable phase transition hysteresis might be applied to a broad range of stimuli-responsive soft materials.

Fig. 1 Schematic illustrations of the fabrication and mechanisms of hydrogels with adaptive memory. (a) The synthetic scheme for the fabrication of the P(NIPAm-co-AAc)-g-MC hydrogel and P(NNPAm-co-AAc)-g-MC. (b) Schemes showing adaptive memory based on tunable phase-transition hysteresis upon environmental pH stimulation. As the solution pH increased from pH 4 to pH 4.5, pH 5, pH 5.5, and pH 6, the LCST of P(NIPAm-co-AAc)-g-MC and P(NNPAm-co-AAc)-g-MC continued increasing due to gradually deprotonated PAAc repeating units. The hydrophobic interactions are enhanced with increasing temperature, resulting in the improvement of thermal hysteresis. At pH 6, the incomplete contracted states of the hydrogels are also maintained when cooling down from T_C1 and T_C2, resulting in a series of relatively small thermal hysteresis loops.

Results and discussion

Tunable thermal hysteresis of fabricated P(NIPAm-co-AAc)-g-MC and P(NNPAm-co-AAc)-g-MC hydrogels

The dual-responsive hydrogel P(NIPAm-co-AAc)-g-MC was prepared by one-pot synthesis (see SI, materials and preparation). MC, NIPAm, AAc and N,N′-methylene benzene (acrylamide) (MBAA) were mixed in distilled water, followed by the addition of ammonium persulfate and N,N,N′,N′-tetramethylenediamine (TEMED). In this reaction, free radicals of the hydroxyl groups on the MC backbone were formed and reacted with the vinyl groups of NIPAm, AAc or MBAA to form covalent bonds between the MC and NIPAm, AAc or MBAA.^34–36 The propagated radicals continue to react with vinyl groups of another NIPAm, AAc or MBAA, which ultimately generates the crosslinked P(NIPAm-co-AAc)-g-MC network. Free radicals of vinyl groups can be also initiated and reacted with MC, NIPAm, AAc or MBAA. To confirm the structure of the P(NIPAm-co-AAc)-g-MC hydrogel, PNIPAm, MC, and the freeze-dried P(NIPAm-co-AAc)-g-MC hydrogel were characterized by Fourier transform infrared spectroscopy (FTIR). The FTIR spectrum of the P(NIPAm-co-AAc)-g-MC hydrogel (Fig. S1) shows characteristic absorptions attributed to PNIPAm, PAAc and MC. The most obvious characteristic of the spectrum for the P(NIPAm-co-AAc)-g-MC hydrogel is the presence of spectral bands derived from the alkoxy group of MC, the amino group of the PNIPAm and the carboxyl groups of the PAAc, which appear at 1050 cm⁻¹, 1540 cm⁻¹ and 1700 cm⁻¹, respectively. The typical OH stretching vibration of MC at 3419 cm⁻¹, the NH stretching vibration of PNIPAm at 3286 cm⁻¹, and the C=O stretching band of PNIPAm at 1637 cm⁻¹ can also be observed. The successful synthesis of the P(NIPAm-co-AAc)-g-MC network is shown by above observations.

To evaluate the tunable thermal hysteresis of the fabricated P(NIPAm-co-AAc)-g-MC hydrogel, the swelling ratio was measured as a function of temperature (Fig. 2). Prior to measuring the swelling ratio, the hydrogel was maintained for enough time (e.g., 5 hours) at a specific temperature to fully reach equilibrium.³⁷ When the P(NIPAm-co-AAc)-g-MC hydrogel was exposed to a solution of pH 4, the hydrogel showed a dramatic volume phase transition close to ∼31 °C (its LCST) during heating (Fig. 2(a)). This transition was found to be fully reversible, but almost no hysteresis is observed. At pH 4.5, the swelling ratio of the P(NIPAm-co-AAc)-g-MC remains unchanged until 38 °C (Fig. 2(b)). Upon heating, an abrupt volume shrinkage of the P(NIPAm-co-AAc)-g-MC starts at 39 °C, which is higher than the temperature for the P(NIPAm-co-AAc)-g-MC hydrogel at pH 4. Some of the PAAc repeating units are deprotonated at pH 4.5, leading to the hydrophilic P(NIPAm-co-AAc) backbone. The hydrophilic backbone promotes the hydration of the polymer chains and inhibits polymer aggregation, resulting in an increase in the LCST. During cooling, the thermal behavior of P(NIPAm-co-AAc)-g-MC is different from that observed in the heating process. Surprisingly, at 40 °C, which is below the LCST observed during heating, the swelling ratio of P(NIPAm-co-AAc)-g-MC is still maintained even after 3 days. This means that the reswelling of P(NIPAm-co-AAc)-g-MC is hindered, which is presumably because the intra/intermolecular interactions of the MC backbones lead to thermal hysteresis in the volume change.¹⁹ Below 38 °C, the hydrogel begins reswelling, and the initial volume is fully recovered after further cooling to 36 °C. This result clearly demonstrates that the P(NIPAm-co-AAc)-g-MC hydrogel has a thermal hysteresis window, which might pave the way for a memory function. As the solution pH is increased to 5, the LCST continues to increase due to the deprotonation of more PAAc repeating units. The thermal hysteresis window at pH 5 also becomes larger (Fig. 2(c)) compared to that at pH 4.5. This is because the hydrophobic association of MC chains intensifies with increasing temperature, resulting in the formation of junctions.³⁸ In contrast, almost no hysteresis is observed for the P(NIPAm-co-AAc) hydrogel upon varying the pH stimulus (Fig. S2). As the pH solution increased to pH 5.5, pH 6 and pH 7, the LCST continued to increase because more and more PAAc repeating units were deprotonated, and the thermal hysteresis window gradually became wider (Fig. 2(d)–(f)). These results demonstrate that the memory function of the P(NIPAm-co-AAc)-g-MC hydrogel is able to switch between memorizing (with hysteresis) and forgetting (no hysteresis) behavior, and the thermal hysteresis window can be adaptively and reversibly adjusted by pH after hydrogel fabrication. The volume change of the P(NIPAm-co-AAc)-g-MC hydrogel can also be affected by pH stimulation (see SI, relationship between pK_a value of the hydrogel and volume change with pH).

Fig. 2 Adaptive memory based on tunable thermal hysteresis of P(NIPAm-co-AAc)-g-MC and P(NNPAm-co-AAc)-g-MC hydrogels upon changing the pH stimulus. Temperature-dependent swelling ratios measured during heating and cooling of the P(NIPAm-co-AAc)-g-MC hydrogel at (a) pH 4, (b) pH 4.5, (c) pH 5, (d) pH 5.5, (e) pH 6, and (f) pH 7. All the solutions are at an ionic strength of 0.01 M. Temperature-dependent swelling ratios measured during heating and cooling of the P(NNPAm-co-AAc)-g-MC hydrogel at (g) pH 4, (h) pH 5, and (i) pH 5.5. All the solutions are at an ionic strength of 0.1 M.

In order to explore the generalization of the proposed strategy and expand the application of adaptive memory materials, we chose the monomer N-n-propyl acrylamide (NNPAm) with a lower LCST (∼21 °C)³⁹ to prepare the poly(N-n-propylacrylamide-co-acrylic acid)-g-methylcellulose (P(NNPAm-co-AAc)-g-MC) hydrogel (see SI, materials and preparation). As seen in Fig. 2(g), the P(NNPAm-co-AAc)-g-MC hydrogel shows almost no hysteresis at pH 4. Similar to the P(NIPAm-co-AAc)-g-MC hydrogel, at pH 5, some of the PAAc repeating units are deprotonated, leading to a hydrophilic P(NNPAm-co-AAc) backbone and an increase in the LCST. During heating, an abrupt volume shrinkage of the P(NNPAm-co-AAc)-g-MC starts at 36 °C, which is higher than the temperature for the P(NNPAm-co-AAc)-g-MC hydrogel at pH 4 (LCST ∼21 °C). During cooling, even at 33 °C, which is below the LCST observed during heating, the swelling ratio of P(NNPAm-co-AAc)-g-MC is maintained (Fig. 2(h)). This means that the reswelling of P(NNPAm-co-AAc)-g-MC is hindered due to thermal hysteretic volume change. Below 33 °C, the hydrogel begins reswelling, and the initial swelling ratio is fully recovered after further cooling to 30 °C. The P(NNPAm-co-AAc)-g-MC hydrogel with a thermal hysteresis window is clearly proved by the above experimental results. As the pH solution was increased to pH 5.5, the LCST continued to increase because more PAAc repeating units were deprotonated (Fig. 2(i)). The thermal hysteresis window at pH 5.5 also became larger compared to that at pH 5. Notably, between pH 5 and pH 5.5, the thermal hysteresis window of the P(NNPAm-co-AAc)-g-MC hydrogel is around 37 °C, which is near body temperature, indicating great potential in biomedical fields.

Furthermore, the thermal hysteresis windows of the above hydrogels may autonomously and adaptively change in response to the surrounding temperature without the pH stimulus being adjusted. When the pH of the solutions was fixed at pH 5, the P(NIPAm-co-AAc)-g-MC hydrogel was exposed to solutions with different ionic strengths. The thermal hysteresis window shifted to the lower temperature region with increasing ionic strength (Fig. 3(a)–(c)), which may be due to the salt-out effect.⁴⁰ Moreover, when the P(NIPAm-co-AAc)-g-MC hydrogel was exposed to a solution of pH 5, and the surrounding temperature was increased to 41 °C followed by a cooling process, the hydrogel exhibited almost no thermal hysteresis (Fig. 3(d)). When the surrounding temperature was increased to 43 °C followed by a cooling process, the hydrogel showed clear thermal hysteresis (Fig. 3(e)). When the surrounding temperature was increased to 47 °C, the hydrogel contracted to become smaller, and the thermal hysteresis window became larger (Fig. 3(f)) compared to that at 43 °C. This is because the hydrophobic association of MC chains intensifies with increasing temperature, resulting in the formation of junctions.³⁸ Similarly, when the P(NIPAm-co-AAc)-g-MC hydrogel was exposed to a solution of pH 6, and the surrounding temperature was increased to 56 °C followed by a cooling process, the hydrogel exhibited almost no thermal hysteresis (Fig. 3(g)). When the surrounding temperature was increased to 58 °C followed by a cooling process, the hydrogel showed clear thermal hysteresis (Fig. 3(h)). When the surrounding temperature was increased to 64 °C, the hydrogel contracted to become smaller, and the thermal hysteresis window became larger (Fig. 3(i)) compared to that at 58 °C. The P(NIPAm-co-AAc)-g-MC hydrogel can maintain a series of relatively small thermal hysteresis loops, which are suitable for adaptively memorizing multiple states. These results demonstrate that the memory function of the P(NIPAm-co-AAc)-g-MC hydrogel is able to switch between forgetting (no hysteresis) and memorizing (with hysteresis) behavior, and the thermal hysteresis window can adapt to the surrounding temperature autonomously.

Fig. 3 Adaptive memory based on tunable thermal hysteresis of P(NIPAm-co-AAc)-g-MC by changing the ionic strength or initial cooling temperature. Temperature-dependent swelling ratios measured during heating and cooling of the P(NIPAm-co-AAc)-g-MC hydrogel at pH 5 upon varying the ionic strength of the buffer solution: (a) 0.3 M, (b) 0.1 M, and (c) 0.01 M. Temperature-dependent swelling ratios measured during heating and cooling of the P(NIPAm-co-AAc)-g-MC hydrogel at pH 5 upon varying the initial cooling temperature: (d) 41 °C, (e) 43 °C, and (f) 47 °C. All these solutions are at an ionic strength of 0.01 M. Temperature-dependent swelling ratios measured during heating and cooling of the P(NIPAm-co-AAc)-g-MC hydrogel at pH 6: (g) 56 °C, (h) 58 °C, and (i) 64 °C. All the solutions are at an ionic strength of 0.01 M.

The molecular mechanism of tunable hysteresis of P(NIPAm-co-AAc)-g-MC

Temperature-dependent FTIR spectra of the P(NIPAm-co-AAc)-g-MC hydrogel (D₂O) during heating and cooling were obtained to confirm the tunable thermal hysteresis and explore the molecular mechanism (see SI, characterization). Fig. 4(a)–(f) show the temperature-dependent FTIR spectra of the P(NIPAm-co-AAc)-g-MC hydrogel in one heating and cooling cycle when the hydrogel was exposed to solutions of pH 4, pH 4.5 and pH 5. In order to eliminate the overlap of the broad ν(OH) band of H₂O at ∼3300 cm⁻¹ with the ν(CH) bands of the hydrogel, D₂O was used to replace H₂O as the solvent.⁴¹ As expected, at pH 4.5 and pH 5, an obvious hysteresis between the heating and cooling processes of the P(NIPAm-co-AAc)-g-MC hydrogel can be observed in the C–H stretching bands at 3000–2960 cm⁻¹. After carefully examining the spectral variations of the C–H stretching bands at 3000–2960 cm⁻¹, we find that all the C–H stretching bands shift slightly to lower frequency in the heating process. The case of the cooling process is opposite to that for heating. The shifts of the ν(C–H) bands can be illustrated by a hydrophobic interaction of the polymer with neighboring water molecules of the solution. The higher the number of water molecules surrounding C–H groups, the higher the vibrational frequency.⁴² The temperature-dependent frequency shifts of ν(CH₃) have been plotted in Fig. 4(g)–(i), in order to quantitatively describe the volume phase transition process. The transition temperature determined by the frequency shift of ν(CH₃) is well matched with those determined from the temperature-dependent swelling ratio in the heating and cooling process, indicating that the tunable hysteresis of the P(NIPAm-co-AAc)-g-MC hydrogel is caused by the varying hydrophobic molecular interactions.

Fig. 4 The molecular mechanism of tunable hysteresis of P(NIPAm-co-AAc)-g-MC. Temperature-dependent FTIR spectra of the P(NIPAm-co-AAc)-g-MC hydrogel (D₂O) during heating at (a) pH 4, (b) pH 4.5, and (c) pH 5, and cooling at (d) pH 4, (e) pH 4.5, and (f) pH 5. Temperature-dependent frequency shifts of ν(CH₃) during heating and cooling processes at (g) pH 4, (h) pH 4.5, and (i) pH 5. (j) Molecular dynamics simulations of the P(NIPAm-co-AAc)-g-MC hydrogel at pH 5. The molecular interaction energies between –CH₃ of PNIPAm and PNIPAm (k) and (n), –CH₃ of MC and PNIPAm chains (l) and (o), and –CH₃ of MC and MC chains (m) and (p) in P(NIPAm-co-AAc)-g-MC hydrogel systems were calculated.

To further explore the hydrophobic molecular interactions of the P(NIPAm-co-AAc)-g-MC hydrogel in the heating and cooling process, all-atom molecular dynamics (MD) simulations were conducted (Fig. 4(j)). The molecular interaction energies between –CH₃ of the MC and MC chains, –CH₃ of the MC and PNIPAm chains, and –CH₃ of the PNIPAm and PNIPAm chains in the P(NIPAm-co-AAc)-g-MC hydrogel systems were calculated (Fig. 4(k)–(p)). All molecular interaction energies were negative, indicating that the interactions between the molecules were attractive.⁴³ The absolute value of the molecular interaction energies at pH 4 is clearly smaller than that at pH 5, indicating stronger hydrophobic molecular interactions at pH 5. At pH 4, the absolute value of the molecular interaction energy between –CH₃ of PNIPAm and PNIPAm chains increases with increasing temperature, and decreases with decreasing temperature (Fig. 4(k)). While at pH 5, the absolute value of the molecular interaction energy between –CH₃ of PNIPAm and PNIPAm chains increases with increasing temperature, and remains unchanged with decreasing temperature (Fig. 4(n)). The molecular interaction energy between –CH₃ of MC and PNIPAm chains follows a similar trend (Fig. 4(l) and (o)). These results show that the molecular interactions between MC and PNIPAm chains, and PNIPAm and PNIPAm chains might also be contributors to the formation of the thermal hysteresis of the P(NIPAm-co-AAc)-g-MC hydrogel. At pH 4, the absolute value of the molecular interaction energy between –CH₃ of MC and MC chains remains unchanged (Fig. 4(m)), probably due to the relatively low temperature range (i.e., 29–35 °C). At pH 5, the absolute value of the molecular interaction energy between –CH₃ of MC and MC chains also increases with increasing temperature, and remains almost unchanged with decreasing temperature (Fig. 4(p)), which supports the concept of intra/intermolecular interactions of the MC backbones leading to thermal hysteresis in the volume change.

Microvalve and hydrogel pattern applications based on adaptive memory

Based on adaptive and tunable thermal behaviors of the P(NIPAm-co-AAc)-g-MC, we tried to exploit a hydrogel microvalve with adaptive memory function. Microvalves play an important role in microfluidic systems, and have versatile functions such as on/off switching, flow regulation, sealing of micro/nano particles and chemical reagents.⁴⁴ This research can be regarded as proof of the concept of the valve application based on adaptive memory. As a demonstration of our concept, a microvalve was formed from a reservoir of dye, a reservoir of water, a channel that connected the reservoirs of dye and water, a piece of polymer film as the “valve” to control the open/close status of the channel, and the P(NIPAm-co-AAc)-g-MC hydrogel loaded with carbon nanotubes for photothermal control (Fig. 5(a), (b), see SI, smart microvalve system). At pH 4, the P(NIPAm-co-AAc)-g-MC hydrogel shows a strict one-to-one correlation between the environment temperature and the swelling ratio. The hydrogel contracted above 32 °C and the blue dye was released via the channel. Below 30 °C, the hydrogel immediately expanded and blocked the channel. At pH 5, multiple states of the swelling ratio can exist in a single environment depending on the history of the hydrogel. The hydrogel expanded at 40 °C during the heating process. Until heating to 48 °C, the channel was opened and blue dye diffused out of its reservoir, while during the cooling process, the hydrogel still remained contracted at 40 °C and the green dye was released via the channel. At pH 6, the thermal memory window shifted to a higher temperature region. The hydrogel expanded at 53 °C during the heating process. Until heating to 64 °C, the channel was opened and blue dye diffused out of its reservoir, while during the cooling process, the hydrogel still remained contracted at 53 °C and the green dye passed through the channel.

Fig. 5 Microvalve and hydrogel pattern applications based on adaptive memory. (a) Tunable thermal hysteresis of P(NIPAm-co-AAc)-g-MC. (b) The design of the microvalve using the P(NIPAm-co-AAc)-g-MC hydrogel. During heating and cooling, experimental images of the microvalve with adaptive memory are shown upon environmental stimulation at pH 4, pH 5 and pH 6. The scale bar is 10 mm. (c) A hydrogel pattern based on P(NIPAm-co-AAc)-g-MC. The scale bar is 40 mm. During heating and cooling, experimental images of the hydrogel pattern with adaptive memory are shown upon environmental stimulation at pH 4, pH 5 and pH 6. All the solutions are at an ionic strength of 0.01 M.

Regarding the potential applications of the developed P(NIPAm-co-AAc)-g-MC hydrogel, we also envisioned hydrogel patterns capable of adaptive information storage (Fig. 5(c), see SI, hydrogel patterns). This research can be regarded as proof of the concept of hydrogel patterning based on adaptive memory. The pH of the operating conditions may be modified (see SI, the modification of pH of operating conditions). Fig. 5(c) shows that at pH 4 the information “NPU” was visualized upon heating at 32 °C and disappeared upon cooling at 30 °C. When the hydrogel pattern was exposed to a solution of pH 5, the memorized information “NPU” was visualized upon heating at 48 °C and still remained upon cooling at 40 °C. At pH 6, the hydrogel pattern is able to adaptively adjust its memory function and shift the thermal memory window to a higher temperature region. The hydrogel pattern shows the information “NPU” upon heating at 64 °C and this is maintained upon cooling at 53 °C. The results demonstrated that the microvalve and hydrogel patterns based on P(NIPAm-co-AAc)-g-MC are able to switch between memorizing and forgetting behavior, and the memory window adapts to environmental stimuli (i.e., pH) autonomously and reversibly.

Smart windows based on adaptive memory

To reduce building energy consumption, thermochromic smart windows have been widely developed to regulate energy exchange in buildings. However, most smart windows are inadequate for varying climates (cold and hot) due to their fixed working temperatures and reliance on continuous electrical energy to maintain optical states.^34,45 This continuous electric consumption paradoxically undermines their energy-saving purpose. Recently, a PNIPAM hydrogel system with additional cross-linking of chitosan and Al³⁺ was prepared, which exhibited a recovery hysteresis in transmittance during cooling.³⁴ The range of hysteresis in transmittance can be tuned within 27∼50 °C by the addition of Al³⁺ ions. However, smart windows based on this material have several limitations. For example: fixed working temperatures are inadequate for varying climates; there is a slow variation in translucency during heating and the transmittance is still more than 50% when the temperature reaches its peak (35–40 °C), which may not effectively block incident light and radiation, and reduce indoor temperature; and the hysteresis windows need the addition of different amounts of salt to be adjusted instead of autonomously changing according to the outside temperature. Hydrogels (e.g., P(NNPAm-co-AAc)-g-MC) with adjustable transition temperatures and tunable memory of the optical state might be an ideal choice for all-weather building temperature regulation in diverse climates and energy conservation of buildings. The optical properties of the fabricated window (Fig. 6(a)) based on P(NNPAm-co-AAc)-g-MC were determined at various temperatures. This research can be regarded as proof of the concept of smart windows based on adaptive memory. As seen in Fig. 6(b) and (c), at pH 4.7 and pH 5.5 the transmittance change for the P(NNPAm-co-AAc)-g-MC window exhibits a similar trend to the swelling ratio change. At pH 4.7, the transmittance changes smoothly over a wide range from 29 to 34 °C in the heating process. The opaque state (0% transmittance) was maintained until 31 °C in the cooling process (Fig. 6(b)), which is beneficial for energy conservation of the building and cooling of the indoor temperature. The transmittance was recovered from 30 °C, and the completely transparent state was observed at 29 °C. At pH 5.5, the transmittance change occurred over a wide range from 37 to 52 °C.

Fig. 6 A smart window based on adaptive memory. (a) A schematic diagram of the preparation of P(NNPAm-co-AAc)-g-MC hydrogel windows with adaptive memory. (b) Photographs of fabricated windows and temperature-dependent transmittance changes measured in the heating and cooling process at pH 4.7. (c) Photographs of the fabricated window in its fully opaque state (i.e., 0% transmittance) and temperature-dependent transmittance changes measured in the heating and cooling process at pH 5.5. (d) Photographs of the smart window in its partially opaque state (i.e., 42% transmittance) and temperature-dependent transmittance changes measured in the heating and cooling process at pH 5.5. All the solutions are at an ionic strength of 0.1 M.

The opaque state (0% transmittance) is reached at 52 °C during heating and maintained until 46 °C during cooling. The transmittance was recovered from 43 °C, and the completely transparent state was observed at 37 °C (Fig. 6(c)). The uniformity of light transmittance for the smart window was tested (see SI, uniformity of light transmittance for smart window). The optical transmittance of the window in its partially opaque (e.g. 42% transmittance) state is reached at 49 °C during heating and maintained without applying any energy, rather than the transparent state being recovered (Fig. 6(d)), indicating suitability for the construction of customizable and adaptive smart windows. This smart window may autonomously and adaptively change its optical transmittance according to the outside temperature without the pH stimulus being adjusted. When the outdoor temperature is very high (e.g., 52 °C), the smart window maintains a very low optical transmittance (e.g., 0% transmittance) and relatively large hysteresis window (e.g., in the range of 11.7 °C) during cooling for effectively reducing the indoor temperature. When the outdoor temperature is high, the smart window maintains a relatively low optical transmittance and relatively small hysteresis window during cooling for reducing the indoor temperature and retaining a comfortable level of sunshine at the same time. The hydrogel smart window with tunable memory provides a foundation for novel smart windows for all-weather temperature regulation in diverse climates and will greatly contribute globally to energy conservation of buildings.

Conclusions

In conclusion, we have developed a series of dual-responsive hydrogels that exhibit adaptive thermal hysteresis via a one-pot synthesis method. To achieve this behavior, we introduced P(NIPAm-co-AAc) or P(NNPAm-co-AAc) chains onto an MC backbone. The memory window adapts intelligently and autonomously to environmental stimuli (i.e., pH), and the range of the thermal hysteresis window can be tuned from approximately 0 °C to 17.6 °C, driven by the stable hydrophobic interaction. In addition, the thermal hysteresis windows adapt to the surrounding temperature autonomously. The P(NIPAm-co-AAc)-g-MC hydrogel can maintain a series of small hysteresis loops, which are suitable for memorizing multiple states. Applications in microvalves, hydrogel patterns and smart windows were successfully demonstrated based on intrinsic hysteresis of the prepared hydrogels. These applications based on P(NIPAm-co-AAc)-g-MC or P(NNPAm-co-AAc)-g-MC involve switching between memorizing and forgetting behavior, with the memory window adapting to environmental stimuli autonomously and reversibly. These hydrogels with adaptive memory could advance the development of sensors, actuators, memory devices, soft matter computers, and soft robots, paving the way for more sophisticated compliant structures and intelligent soft robotic systems.

Author contributions

Xuan Zhang: writing – review & editing, writing – original draft, funding acquisition, conceptualization. Zichao Wang: methodology, investigation, data curation. Xuehua Zhou: investigation, data curation. Mingze Liu: writing – review & editing. Xuefeng Zhu: formal analysis. Mingchao Zhang: review & editing. Xuzi Yang: review & editing. Yinglai Hou: writing – review & editing, software, conceptualization. Yuzhang Du: data curation. Jie Kong: writing – review & editing, supervision, conceptualization.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). The supplementary information includes materials and methods, characterization, smart microvalve system, hydrogel patterns, the modification of pH of operating conditions and uniformity of light transmittance for smart window. See DOI: https://doi.org/10.1039/d5mh01416f.

Acknowledgements

This work was financially supported by the National Natural Science Foundation of China (52373120, 52003220), National Science Fund for Distinguished Young Scholars (2025034), Aoxiang rising star of Northwestern Polytechnical University (24SH0201283, 24GH0201283) and Key Research and Development Program of Shaanxi Province (2023-YBGY-193).

Notes and references

1. X. Zhang, L. Chen, K. H. Lim, S. Gonuguntla, K. W. Lim, D. Pranantyo, W. P. Yong, W. J. T. Yam, Z. Low, W. J. Teo, H. P. Nien, Q. W. Loh and S. Soh, Adv. Mater., 2019, 31, 1804540.
2. A. Ghosh, G. Koster and G. Rijnders, Adv. Funct. Mater., 2016, 26, 5748–5756.
3. L. Shao, X. Xu, Y. Liu and Y. Zhao, ACS Appl. Mater. Interfaces, 2023, 15, 35272–35279.
4. C. Yu, H. Guo, K. Cui, X. Li, Y. Ye, T. Kurokawa and J. P. Gong, Proc. Natl. Acad. Sci. U. S. A., 2010, 117, 18962–18968.
5. Y. Xia, Y. He, F. Zhang, Y. Liu and J. Leng, Adv. Mater., 2021, 33, 2000713.
6. A. Lendlein and O. E. C. Gould, Nat. Rev. Mater., 2019, 4, 116–133.
7. S. Choudhury, A. Joshi, V. S. Bagher, G. K. Ananthasuresh, S. Asthana, S. Homer-Vanniasinkam and K. Chatterjee, J. Mater. Chem. B, 2024, 12, 5678–5689.
8. Q. Zhao, H. J. Qi and T. Xie, Prog. Polym. Sci., 2015, 49, 79–120.
9. W. Yuan, J. Zhou, K. Liu, X. Li, W. Xu, H. Song, G. Shan, Y. Bao, Q. Zhao and P. Pan, ACS Macro Lett., 2020, 9, 588–594.
10. H. Xie, C. Cheng, X. Deng, C. Fan, L. Du, K. Yang and Y. Wang, Macromolecules, 2017, 50, 5155–5164.
11. X. Tao, Nature, 2010, 464, 267–270.
12. Y. Shi, C. B. Cooper, T. Nogusa, J. Lai, H. Lyu, M. Khatib, C. Xu, L. Michalek and Z. Bao, Matter, 2024, 7, 2108–2124.
13. S. Yang, Y. He, Y. Liu and J. Leng, J. Mater. Chem. C, 2020, 8, 303–309.
14. X. Liu, M. Gao, J. Chen, S. Guo, W. Zhu, L. Bai, W. Zhai, H. Du, H. Wu, C. Yan, Y. Shi, J. Gu, H. Qi and K. Zhou, Adv. Funct. Mater., 2022, 32, 2203323.
15. A. Joshi, S. Choudhury, A. Majhi, S. Parasuram., V. S. Baghel, S. Chauhan, S. Khanra, D. Lahiri and K. Chatterjee, Biomater. Sci., 2025, 13, 4706.
16. K. Itano, J. Choi and M. F. Rubner, Macromolecules, 2005, 38, 3450–3460.
17. K. E. Secrist and A. J. Nolte, Macromolecules, 2011, 44, 2859–2865.
18. J. Michalska-Walkowiak, B. Förster, S. Hauschild and S. Förster, Adv. Mater., 2022, 34, 2108833.
19. D. Kim, H. Kim, E. Lee, K. S. Jin and J. Yoon, Chem. Mater., 2016, 28, 8807–8814.
20. J. L. Weidman, R. A. Mulvenna, B. W. Boudouris and W. A. Phillip, J. Am. Chem. Soc., 2016, 138, 7030–7039.
21. M. Annaka and T. Toyoichi, Nature, 1992, 355, 430–432.
22. I. Franck, T. Toyoichi and K. Etsuo, Nature, 1991, 349, 400–401.
23. K. Zhang, X. Feng, C. Ye, M. A. Hempenius and G. J. Vancso, J. Am. Chem. Soc., 2017, 139, 10029–10035.
24. Z. Jiang, B. Diggle, I. C. G. Shackleford and L. Connal, Langmuir, 2010, 26, 9224–9232.
25. Z. Jiang, B. Diggle, I. C. G. Shackleford and L. A. Connal, Adv. Mater., 2019, 31, 1904956.
26. H. G. Schild, Prog. Polym. Sci., 1992, 17, 163–249.
27. L. Sambe, V. R. de La Rosa, K. Belal, F. Stoffelbach., J. Lyskawa, F. Delattre, M. Bria, G. Cooke, R. Hoogenboom and P. Woisel, Angew. Chem., Int. Ed., 2014, 53, 5044–5048.
28. V. R. de la Rosa, W. M. Nau and R. Hoogenboom, Org. Biomol. Chem., 2015, 13, 3048–3057.
29. A. Saha, S. Manna and A. K. Nandi, Biomaterials, 2004, 25, 3005–3012.
30. A. Saha, S. Manna and A. K. Nandi, Soft Matter, 2009, 5, 3992–3996.
31. M. R. Berber, H. Mori, I. H. Hafez, K. Minagawa, M. Tanaka, T. Niidome, Y. Katayama, A. Maruyama, T. Hirano, Y. Maeda and T. Mori, Polym. J., 2010, 114, 7784–7790.
32. B. Ray, Y. Okamoto, M. Kamigaito, M. Sawamoto, K. Seno, S. Kanaoka and S. Aoshima, Polym. J., 2005, 37, 234–237.
33. L. Sambe, Victor R. de La Rosa, K. Belal, F. Stoffelbach, J. Lyskawa, F. Delattre, M. Bria, G. Cooke, R. Hoogenboom and P. Woisel, Angew. Chem., Int. Ed., 2014, 126, 5144–5148.
34. B. Liu, J. Liu and Y. Yu, Eur. Polym. J., 2022, 162, 110929.
35. D. Roy, M. Semsarilar, J. T. Guthrie and S. Perrier, Chem. Soc. Rev., 2009, 38, 2046–2064.
36. V. N. Kislenko, J. Colloid Interface Sci., 1999, 209, 136–141.
37. L. Dong and H. Jiang, Soft Matter, 2007, 3, 1223–1230.
38. L. Li, P. M. Thangamathesvaran, C. Y. Yue, K. C. Tam, X. Hu and Y. C. Lam, Langmuir, 2001, 17, 8062–8068.
39. C. Zhao, Z. Ma and X. X. Zhu, Prog. Polym. Sci., 2019, 90, 269–291.
40. Y. Xu and L. Li, Polymer, 2005, 46, 7410–7417.
41. S. Sun, J. Hu, H. Tang and P. Wu, J. Phys. Chem. B, 2010, 114, 9761–9770.
42. P. Schmidt, J. Dybal and M. Trchová, Vib. Spectrosc., 2006, 42, 278–283.
43. Y. Yang, W. Wu, H. Liu, H. Xu, Y. Zhong, L. Zhang, Z. Chen, X. Sui and Z. Mao, J. Mol. Graphics Modell., 2010, 97, 107554.
44. W. Wang, P. F. Li, R. Xie, X. J. Ju, Z. Liu and L. Y. Chu, Adv. Mater., 2022, 34, 2107877.
45. G. Chen, K. Wang, J. Yang, J. Huang, Z. Chen, J. Zheng, J. Wang, H. Yang, S. Li, Y. Miao, W. Wang, N. Zhu, X. Jiang, Y. Chen and J. Fu, Adv. Mater., 2023, 35, 2370139.

---

Footnote

† These authors contributed equally to this work.

---