Introduction

Supramolecular polymer gels are a class of soft matter possessing 3D polymeric networks, formed through transient covalent or non-covalent bonding such as hydrophobic interactions, electrostatic attraction, hydrogen bonds, π-π interactions, and metal-ligand coordination bonding, using moieties such as ureido-pyrimidinone, terpyridines, crown-ethers, and catechols (Dong et al. 2011, 2012; Kurth 2008; Lehn 1995; Lehn et al. 1992; Mingos 2004; Ott et al. 2012; Yan et al. 2012; Zerkowski et al. 1992). Among them, hydrogen bonds and π-π interactions are valuable tools in engineering supramolecular systems, especially in non-aqueous solvents (Sijbesma et al. 1993). Nowadays, many different types of chemically crosslinked hydrogels and non-covalently crosslinked supramolecular gels have been developed to respond to external stimuli such as pH, temperature, or ions (Faure et al. 2013; Sangeetha and Maitra 2005; Vatankhah-Varnoosfaderani et al. 2013; Zhang et al. 2012). Among others, supramolecular bonding facilitates the design of self-healing gels that can spontaneously recover their strength after fracture, as self-healing allows for the formation of adaptive smart materials for a variety of applications such as in the biomimetic field (Phadke et al. 2012; South and Lyon 2010). Therefore, it is important to develop strategies to design and construct strong supramolecular gels with outstanding mechanical behavior as well as self-healing properties (Cong et al. 2013). As previous articles have shown, supramolecular gels are usually weak, which can be improved by micro- or nano-fillers (Chunder et al. 2010; Fox et al. 2012; Mandal et al. 2010). While various sheet-like clays such as Laponites or Montmorillonites have previously been used, carbon nanotubes (CNTs) and graphene oxide (GO) are used nowadays additionally as active fillers capable of improving gel properties (Whiteside et al. 2013).

Graphene nanosheets have higher surface-to-volume ratios than single-walled CNTs or even multi-walled CNTs owing to the inaccessibility of the inner nanotube surfaces to polymer molecules (Stankovich et al. 2006). Due to its atomic thickness and lower density, GO also has a higher surface area than nanoclays. This potentially makes graphene nanosheets more favorable for altering the mechanical properties (Zhang et al. 2011). Recently, we reported a copolymer of p(N-isopropylacrylamide-co-dopamine methacrylate) (NIDO5% with 5 wt.% dopamine methacrylate) fabricated by free-radical polymerization (Vatankhah-Varnoosfaderani et al. 2014). In the presence of sodium tetrahydroborate (in non-aqueous media), this copolymer leads to a foamed gel, which is not self-healing except during the absorption of water molecules from the environment. However, this change in behavior is achieved by decreasing the storage modules to only 22 % of that of the anhydrous sample, which limits its application (Vatankhah-Varnoosfaderani et al. 2013).

In this work, we show that after the incorporation of reduced graphene oxide (RGO) into such an organogel, a more rigid and “solid-like” gel is formed, while the fast self-healing ability of the prepared composite gel was almost as good as that of wet organogels or hydrogels of the same polymer. At the same time, RGO improved its mechanical properties in wet condition. Whereas the gelation of these nanocomposites is reversible by adjusting the pH, they can also swell up in water up to 250 wt.%, which induces thermal sensitivity, i.e., the appearance of a lower critical solution temperature (LCST). Therefore, these nanocomposite gels are sensitive to two external stimuli. In the future, systems with multi stimuli-responsiveness and self-healing properties will open up new applications and much effort has already been devoted to the design of a synthetic self-healing material inspired by mussel byssal proteins and threads (Brubaker et al. 2010).

Experimental

The synthesis of the copolymer of N-isopropylacrylamide (NIPAM) and dopamine methacrylamide (DMA) (NIPAM-co-DMA) with a DMA content of 5 %, abbreviated as NIDO5%, has been previously described (Vatankhah-Varnoosfaderani et al. 2013). Gel permeation chromatography and vapor pressure osmometry were used to deduce a molecular weight M n of ≈2800 g/mol and M w /M n  ≈ 1.9 for the material (Vatankhah-Varnoosfaderani et al. 2013, 2014).

Graphene oxide (GO) was prepared by a modified Hummer’s method (Hirata et al. 2004). Graphite flakes were added to a round-bottom flask containing 120 ml of H2SO4 under stirring for 1 h. KMnO4 aqueous solution was titrated to the mixture at an interval of 20 ml. The mixture was maintained at 40 °C for 5 h and then 150 ml of deionized water was added to the mixture. Additionally, H2O2 solution was added dropwise to the mixture within 30 min and then maintained for 24 h. After exfoliated GOs were separated by centrifuge, they were put into the dialysis tube for neutralization. Finally, the GOs were dried by freeze-drying equipment.

The procedure of preparing typical reduced GO (RGO) sheets is described as follows (Dang et al. 2012). The aqueous suspension of GO (10 mg/ml) was diluted with dimethyl formamide (DMF) to a concentration of 2 mg/ml and sonicated in an ultrasonic bath (Jeiotech UC-10, 200 W) for 1 h to create a homogenous suspension of GO in DMF/water (80:20 v/v). The hydrazine reduction was performed by adding 5 ml of hydrazine monohydrate (98 %) to 100 ml of the graphene suspension, which was stirred at a desired temperature (40 °C) in an oil bath for 24 h. The resulting reduced GO samples were filtered and washed with copious amounts of DMF to remove excess hydrazine. The RGO filter cakes were redispersed in DMF by sonication for 2 h with the temperature of the sonication bath being maintained at less than 30 °C.

Fourier transform infrared spectroscopy (FTIR, Perkin-Elmer Spectrum One, USA) was used to study the functionalization. Native and composite dried powdered gels were molded into disks using KBr. A micro-Raman spectrometer (Nanofinder 30) with an argon ion laser at an excitation wavelength of 632.8 nm and an infrared spectrometer (Tokyo Instrument, Inc.) was employed.

X-ray diffraction (XRD) experiments were performed at room temperature to study the crystalline structure of materials using an XRD system (Philips X’Pert-MRD, the Netherlands) using CuKα radiation (X-ray wavelength λ = 1.5406 Å) under normal laboratory conditions. The chemical compositions and crystallographic structures of RGO and the native and composite gels were recorded by varying the angle following Bragg’s law, which is satisfied by the d-spacing in polycrystalline materials. Plotting the angular positions and intensities of the resultant diffracted peaks of the radiation produces a pattern which is characteristic of the sample, where a mixture of different phases are present.

Transmission electron microscopy (TEM, JEOL (Japan)/JEM-2010 operated at 200 kV) was used to demonstrate the morphology and 2D structural information of the RGO and NIDO-RGO nanocomposites. The samples were purified by repeated washing and centrifugation to remove free polymer chains prior to this measurement. For NIDO-RGO nanocomposites, the blend of NIDO and RGO in DMF was prepared and filtered through 0.2 μm PTFE membrane. The solid residue obtained was repeatedly washed with an excessive amount of DMF through dialysis tube to ensure satisfactory washing of excessive polymer. Finally, the washed RGO was dispersed in DMF for sample preparation.

For assessing the swelling performance, distilled water was added to the vials containing the gels. After reaching equilibrium for 1 day at room temperature, the excess water was taken out and the surface of the gel and the vials were dried by nitrogen gas prior to the weighing the gels. The amount of swelling S was calculated as

$$ S=\left({w}_s-{w}_n\right)/{w}_n. $$
(1)

Herein, w s and w n represent the weights of swollen gel and non-swollen gel (organogel), respectively.

The mechanical/rheological properties of the gels were evaluated using a rheometer (Malvern Kinexus Pro) with a cone and plate geometry (20 mm diameter rotating top plate). The bubbles, stemming from the decomposition of NaBH4, were removed by tapping and centrifuging. For the dry samples, humidity was carefully excluded by drying all components in the sample chamber. The humid samples were exposed to ambient humidity for several days prior to rheological testing. Oscillatory shear testing of the gels as a function of frequency was performed at a constant strain of γ 0 ≈ 1 %. Recovery tests or self-healing tests were performed by increasing the deformation stepwise from γ 0 = 0.1 to 300 % (ω = 10 s−1), i.e., a so-called strain sweep, followed immediately by a time-sweep under linear conditions (γ 0 < 1 %, ω = 10 s−1), while monitoring the recovery of the storage modulus, G′, and loss modulus, G″. Swollen samples were tested under water saturated conditions, preventing water loss during testing by a solvent trap.

Unfortunately, it is not possible to do elongational experiments on this kind of gel. Stadler et al. (2013) found that it is not possible to do good tensile tests on samples with a modulus below 5000 Pa. The sample 10%NIPAM-5%AUTMAB (Stadler et al. 2013) was so soft that it was almost impossible to load it, and these samples were very ductile due to the chemical structure. The hydrogels and organogels in this article are lower in modulus and, furthermore, the supramolecular bonding means that stretching them will start stress-relaxation processes by relaxation of network segments, which are enabled by temporarily opened supramolecular bonds. Menyo et al. (2013) proved this by relaxation tests on catechol- and catechol-derived functionalities. These are the reasons why, to the best of our knowledge, no publications exist on tensile properties of supramolecular soft gels (G < 5000 Pa).

Photographs were taken with a Canon EOS 5D Mk II equipped with a Canon 24–70 mm/1:2.8 lens at 70 mm from minimum focal distance (ca. 38 cm) so as to ensure constant size and lighting conditions. Lighting was supplied by a steady light source placed above the camera.

Results and discussion

In the supramolecular polymer NIDO5%, both monomers have N-containing functionalities (Fig. 1a), accepting protons from functional groups of the GO sheets. The network formation between PNIPAM, which is the major monomer of the copolymer, and functional groups on the graphene sheets has been studied previously (GhavamiNejad et al. 2014; Lee et al. 2011; Sahoo et al. 2006). Moreover, catechols can bond covalently to organic surfaces or attack epoxy rings and hydroxyl groups, which has been reported as a green method to reducing the functional groups on GO surface (Cheng et al. 2013a, b; Kang et al. 2011). However, to design smart supramolecular systems through metal-ligand interactions, these catechols should not be crosslinked with GO, as due to the high concentration of oxygen-containing moieties, the DOPA groups react (Cheng et al. 2013b) Thus, GO was reduced by hydrazine monohydrate according to the procedure of Dang et al. (2012), consequently, eliminating most but not all of the functional groups on the GO surface.

Fig. 1
figure 1

a Proposed interactions between RGO and polar groups of NIDO5% through hydrogen bonding and Reversible boron-dicatechol complexation. b Assembly mechanism of NIDO5%/RGO based on hydrogen bonding and complexation

Full size image

Although hydrazine reduction is effective for the removal of epoxy and hydroxyl groups on the basal plane, some carboxylic and carbonyl moieties on the edges of graphitic skeleton remain intact (D'Souza and Kadish 2012), which were proven to form hydrogen bonds with –NH-groups and –C = O groups under appropriate conditions (Fig. 1a), which fosters the self-assembly of graphene sheets under suitable conditions (GhavamiNejad et al. 2014; Lee et al. 2011; Sahoo et al. 2006). Furthermore, NIDO5% contains aromatic catechol groups, which should be ideal for interacting with the RGO through π-π stacking. Previous works demonstrated that complexation of a multifunctional catechol polymer with bifunctional boron compounds gave rise to self-healing hydrogels, as well as an organogel exhibiting high stability at alkaline pH and low stability under acidic conditions (Fig. 1a) (He et al. 2011; Vatankhah-Varnoosfaderani et al. 2013). This work demonstrates that the dominating gelation mechanisms are reversible interactions between boron compounds and catechol groups, on one hand, and between RGO sheets and polymer chains, on the other hand (Fig. 1b).

Gel preparation

The procedure for preparing typical 1 wt.% RGO composite gels is described as follows. Under nitrogen, 100 μl of RGO dispersion was added to an anhydrous copolymer solution in DMF and shaken vigorously for 10 s, followed by the addition of 0.5 eq. NaBH4 in DMF relative to catechol-concentration to start the reaction boron-dicatechol complexation. This was followed by the generation of H2 gas, rising until the progressing gelation trapped it, thus forming a foamed composite gel (a movie is available in the SI). Based on macroscopic observation, it is clear that ionic compounds such as NaBH4 are essential for gelation, as without them NIDO5% is adsorbed on RGO (by hydrogen bonding) due to the very low amount of RGO sheets. It has been reported by many researchers that at a high ratio of graphene to polymer, much higher than the concentration used in this study, the gelation can happen for the blend of polymer and graphene sheets as well (Bai et al. 2011).

TEM

The transmission electron microscopy (TEM) images of the RGO sheets (Fig. 2a) and blend of NIDO5% and RGO (Fig. 2b) confirmed the bonding of the copolymer with RGO. As shown in the TEM images, the RGO sheets are very thin (Fig. 2a) but interaction with NIDO5% (Fig. 2b) affords clear black spots on the RGO surface, indicating that edges and basal planes of the RGO sheets are decorated with polymer chains bound by H-bonding and π-π interactions, as discussed before (GhavamiNejad et al. 2014; Lee et al. 2011; Sahoo et al. 2006). The single absorbance peak appeared before and after blending the copolymer with RGO at 280 nm in UV–vis spectroscopy (Figure SI1), and the absence of additional peaks at wavelengths related to catecholic bonding showed that the catechol groups do not play a role in this attachment. The blend (Fig. 2b) was washed several times prior to TEM, as described in the experimental section. Although, it is not directly possible to quantify the amount of NIDO 5 % adsorbed on the surface an impossible task as no permanent bond exists and, hence, the adsorbed amount depends heavily on the sample preparation method—the presence of these spots is a strong indicator of the abovementioned interactions.

Fig. 2
figure 2

TEM images of a RGO and b NIDO5% + RGO blend

Full size image

FTIR

Possible interactions between RGO nanosheets and NIDO5% chains were further clarified by analyzing the FTIR spectra of the RGO powder, native, and nanocomposite gels after solvent removal (Fig. 3). The FTIR spectrum of RGO clearly shows the existence of carboxylic and carbonyl moieties on the edges of reduced GO, which shows the bands at 1722 and 1630 cm−1, respectively. The former is most often related to the –C = O stretching motions of COOH groups situated at the edges of the RGO sheets. In the dried NIDO5% gel, FTIR spectrum bands at 3300 and 1525 cm−1 indicate stretching of the N-H bond, and the absorption bands detected at 1640 and 1460 cm−1 are characteristic of the –C = O groups of the polymer and CH2 structural bending in alkanes, respectively (Park et al. 2012; Vatankhah-Varnoosfaderani et al. 2014). For the RGO/NIDO5% gel, these bands are slightly shifted to lower wavenumbers and weaker, suggesting intertwined networks in the organogel through H-bonding of oxygen-containing groups of RGO sheets with polar groups of copolymer chains (Cong et al. 2013; Jeon et al. 2013).

Fig. 3
figure 3

FTIR spectra of RGO powder, dried native gel, and composite gel

Full size image

RAMAN

Raman spectra of GO, RGO, and blend of NIDO5% and composite gel have been shown in Fig. 4. As can be clearly seen, the Raman spectrum of graphene oxide exhibited two prominent peaks at 1350 and 1595 cm−1, corresponding to the D and G bands, with a D/G-ratio of 1.03. After hydrazine reduction, the G-peak of RGO is gradually shifted to 1582 cm−1 and the D/G-ratio of the RGO increased slightly (ratio D/G = 1.07). The shift of G band and increase in the D/G-ratio are attributed to the restoration of sp2 as new graphitic domains. The RAMAN spectrum of the composite gel is nearly identical to that of RGO. However, in comparison with the RAMAN spectrum of RGO, the intensity ratio of D band to G band has increased. The increase of intensity ratio of the D band and G band indicates increased disorderedness (i.e., sp3 character) because of hydrogen bonding with polymer chains (Kundu et al. 2012). This conclusion is in good agreement with the results given by FTIR experiments.

Fig. 4
figure 4

Raman spectra of GO, RGO, and dried composite gel

Full size image

XRD

The XRD patterns of GO, RGO, NIDO5%, and NIDO5%/RGO (Fig. 5) were used to determine the degree of exfoliation of the graphene sheets (Miller et al. 2010) The characteristic XRD peak of the pure GO sheets appeared at 2θ = 11.26°, corresponding to an interplanar spacing of 7.85 Å, which was significantly higher than the value of 3.35 Å of graphite (Palser 1999; Tung et al. 2011). Hydrazine reduction broadened this peak and shifted it to 2θ ≈ 24.6°, corresponding to 3.61 Å (Dang et al. 2012). This shift for RGO was attributed to the removal of many oxygen functional groups from the GO sheet surface (Dang et al. 2012). The polymer shows a diffraction peak at 2θ = 19.70°, corresponding to amorphous PNIPAM (GhavamiNejad et al. 2014). For the nanocomposite gel, the XRD pattern only shows the polymer without any obvious RGO diffraction peak, clearly demonstrating the uniform dispersion of RGO nanosheets in the gels (Huang et al. 2012; Shen et al. 2012; Sun and Wu 2011). However, it is also clear that the diffraction peak position is changed and has become asymmetric. A comparison between the peak for NIDO5% and the composite gel reveals an almost perfect superposability below 2θ = 20°, while a second peak at around 2θ = 22.2° becomes visible (dotted red line). Deducting the solid leaves a green line (NIDO5% scaled to the blend between 15 and 20°) from the diffraction peak of the blend. This suggests a significant fraction of NIDO5% with more free volume, which was attributed to the proximity to the RGO sheets, where steric hindrance and surface interactions prevent closer average distances. Alternatively, it could also be interpreted as intercalated RGO galleys, which, however, is approximately suggestive of the same structure after NIDO5% has penetrated into the RGO galleys to some degree and, thus, severely distorted the structure while retaining some more free volume than a normal polymer. As the amorphous interpolymer distance is around 4 Å, which is approximately the same as a distorted graphene stack, both explanations are possible and both prove the presence of intercalation and possible exfoliation. The latter explanation is not considered likely, however, as only 1 wt.% of RGO is in the gel, which would lead to a significantly weaker signal should stacked RGO still be present. Furthermore, the peak position is shifted to lower 2θ angle.

Fig. 5
figure 5

XRD patterns of GO powder, RGO powder, dry NIDO5% gel, and dry RGO/NIDO5% gel

Full size image

Rheology and self-healing behavior of dry and hydrous organogels

Frequency sweeps in the linear-viscoelastic regime (γ 0 = 1–5 %, Fig. 6a) compare storage moduli G′(ω) for a regular gel and a corresponding RGO composite gel in dry and wet states of otherwise identical composition. The wet state was created by exposing the gel to ambient humidity for ≈24 h, which leads to a water content of ≈1 wt.% (Vatankhah-Varnoosfaderani et al. 2013). In DMF, catechol functionalities form a quasi-permanent dicomplex with B3+ ions stemming from NaBH4 at pH ≈ 8. This leads to crosslinking of two catechols, which can form either an intra- or interchain bond. While in the presence of water, the boron-dicatechol dicomplex is in equilibrium with H2O, which changes the catechol–borate covalent bonds from irreversible to reversible (Vatankhah-Varnoosfaderani et al. (2013)). Upon incorporation of 1 wt.% RGO in dry state, the gel and composite gel show a G′(ω)-increase from 1200 Pa to 1850 Pa (Fig. 6b). The difference in modulus is related to complex strength and bonding probability, highlighting the contribution of RGO to the gel network. RGO enhances the physical properties by additional crosslinking points provided by hydrogen bonding between the RGO and polymer chains. In wet condition, the reduction of storage modulus G′(ω) of the samples without RGO is 78 % compared to the anhydrous sample, while in the presence of RGO sheets, this reduction is much weaker (12 %). The loss modulus G″(ω) shows an increase for the composite gel upon exposure to humidity, which is not found for the native gel (Fig. 6b). However, this is easily explained by the significant reduction of stiffness (G′(ω)) for the latter. The influence of water can also be seen by the phase angle δ, which increases from 1° by about 0.5° upon addition of humidity for the native gel. The addition of RGO also slightly increases the δ value from 1° to 2.5° with a comparable influence of H2O. As catechols are pH-responsive, lowering the pH below about 7.5 would lead to a breakdown of network structure, which can be rebuilt by increasing the pH above 7.5 again. This issue was investigated and discussed in detail by us earlier (Vatankhah-Varnoosfaderani et al. 2014) As the influence of RGO on these properties is marginal, the pH responsiveness is not investigated in this work.

Fig 6
figure 6

a Frequency sweeps of RGO composite gel compared to a gel without RGO in dry and wet states and b comparison of the data at ω = 1 s−1

Full size image

The nonlinear behavior of the composite gel samples is significantly different from that of the native gels (Fig. 7). As previously discussed, the dry native gel shows a rather low onset of nonlinearity around γ 0 = 3 % (Vatankhah-Varnoosfaderani et al. 2013). The main nonlinearity limit for the dry composite gel is γ 0 ≈ 1 %, which is the typical effect of fillers on the rheological behavior (Anderson and Zukoski 2009; Cassagnau 2003; Choong et al. 2013). When considering G′(γ 0) and G″(γ 0), multiple relaxations can be seen. For the native dry gel, clear curvature changes around 20 and 150 % deformation are obvious in G′(γ 0), while only the latter can be seen in G″(γ 0) as a peak, whereas the former shows as the upturn of G″(γ 0) (Vatankhah-Varnoosfaderani et al. 2013). For the dry composite gel, these transitions are also visible at roughly the same deformations. However, another curvature change (nonlinearity transition) at γ 0 ≈ 3 % is obvious in G′(γ 0), which appears in G″(γ 0) at around γ 0 = 1 % and is clearly related to the influence of RGO. The influence of water lies not only in the lower modulus but also in the higher nonlinearity limit due to the change of nature of the boron-dicatechol bond from permanent to transient (Vatankhah-Varnoosfaderani et al. 2013). Comparison of the hydrous native and covalent gels reveals that the nonlinearity limit decreases upon addition of RGO from γ 0 ≈ 25 % to ≈12 %. Like the dry gels, the hydrous gels exhibit a type III-behavior (Hyun et al. 2002; Sim et al. 2003): G′(γ 0) decreases with rising γ 0, while G″(γ 0) goes through a maximum slightly beyond the nonlinearity limit. Comparison of the native and composite gel samples, in case of the dry composite gel, reveals a clear second process around γ 0 = 100 % from a shoulder in G′(γ 0), also responsible for the G″-peak, while the onset of nonlinearity is significantly earlier. This indicates a two-step relaxation, supposedly caused initially by breakdown of the network structure by RGO-polymer interactions, then followed by the pure gel structure. Adding humidity smears out these two steps, leading to a broader nonlinearity transition than the native, hydrous gel.

Fig. 7
figure 7

Strain sweeps—comparison of G′(γ 0) and G″(γ 0) for the anhydrous and hydrous native (Vatankhah-Varnoosfaderani et al. 2013) and composite gels

Full size image

The dynamic nature of the gel structure was tested by increasing γ 0 from 0.1 to 300 % (ω = 10 s−1), followed by a 20 min recovery to assess the self-healing (γ 0 ≈ 0.5 %, ω = 10 s−1, Fig. 8)). In comparison to the native gel data, the self-healing speed is comparably fast, except for the dry composite gel (Vatankhah-Varnoosfaderani et al. 2013). The loss in modulus due to the shear of γ 0 = 300 % is slightly larger (80 vs. 63 %, (Fig. 8a), which was attributed to the influence of the nanofiller and the effects on rheological behavior discussed above. For the hydrous composite organogel (Fig. 8b), slight damage is found, which is obvious as a reduction of G′(t) of about 14 % in comparison to the value before intensive shear. In comparison to the native hydrous gel, this slight decrease in self-healing ability confirms the strong effect of water on the bonding of boron-dicatechol systems (Vatankhah-Varnoosfaderani et al. 2013). The similarly fast self-healing ability of native and composite gels, both in dry and wet states, proves that the destructuring in both cases occurs through the same process—cleavage of the boron-dicatechol bonds. This suggests that the connections of polymer chains to the RGO sheets in the composite gels are not affected by the self-healing behavior.

Fig. 8
figure 8

Strain sweep with subsequent self-healing test for nanocomposite organogel (γ 0 max = 300 %): a anhydrous and b hydrous organogel

Full size image

The self-healing ability was also tested macroscopically by cutting the wet composite gel apart and rejoining it (Fig. 9), which starts the self-healing. Within about 30 s, the pieces have self-healed back together so that their strength matches the rest of the gel, confirming the above rheological results.

Fig. 9
figure 9

Photographs showing the self-healing and recycling properties of wet RGO/NIDO5% gel: a gel before deformation, b cracked gel, c merged gel, and d merged gel under stretching

Full size image

Properties of swollen samples

Considering the requirement for water absorption into gels in many applications (Friedrich et al. 2011b; Friedrich et al. 2011a; Friedrich et al. 2010; Hashmi et al. 2012; Lu et al. 2014), the role of RGO in the swelling behavior of the gels was investigated by immersion of the gels with and without RGO into deionized water to reach equilibrium swelling. When the organogels are immersed into deionized water, solvent exchange occurs within the gels. As shown in Fig. 9, both samples maintained their shape and, thus, gel characteristics at an equilibrium water content of 280 and 265 % water by weight for the native and composite gels, respectively (Fig. 9), are retained, i.e., a solid content of 5.7 wt.% of polymer is possible, which is remarkable, as no hydrogel is formed for NIDO5% concentrations below approximately 10 wt.% in water.

In comparison, the native gels have a slightly higher equilibrium swelling than corresponding RGO gels (280 vs. 265 %), which can be interpreted as the consequence of the NIDO5%-RGO interactions stiffening the gel and, thus, increasing the barrier for additional water absorption due to countering osmotic pressure. For the following measurements, however, the water adsorption was set to 250 %, so as to ensure comparability of the samples. The slight reddening of the swollen samples (Fig. 10) is the consequence of conversion of catechols to quinones and semiquinones and an unavoidable consequence of the presence of H2O in the system (and in general, for any system containing catechols) (Vatankhah-Varnoosfaderani et al. 2013).

Fig. 10
figure 10

Photographs showing the composite and native gels before and after swelling. From left to right: native gels (anhydrous and swollen) and composite gels (anhydrous and swollen)

Full size image

The dynamic properties of the swollen gels show that the gel behavior is unaffected by the swelling, which can be seen from the elastic modulus being constant and about one order of magnitude higher than the loss modulus (see Fig. 11). Whereas the neat gels display a reduction in G′ from ≈1600 to ≈40 Pa upon swelling (by a factor of ≈40), the nanocomposite gels under the same conditions show a reduction in modulus by a factor of 15 (from ≈1850 to ≈120 Pa), which is in accordance with the trends described in Fig. 7, i.e., a significantly smaller effect of a small amount of humidity on the modulus for the composite gel. Classical crosslinking dynamics for polymer solutions and gels (Koike et al. 1995; Watanabe et al. 1998) suggest a modulus reduction of between 3.5 and 5.1 depending on the dilution exponent α in

$$ {\mathrm{M}}_{\mathrm{e}}\left(\upvarphi \right)={{\mathrm{M}}_{\mathrm{e}}}^{bulk}/{\upvarphi}^{\upalpha} $$
(2)

where φ is the fraction of materials. While the changes observed are significantly larger than predicted, especially for the native gel, the change in modulus is much more severe, which raises the question as to why such a difference is observed.

Fig. 11
figure 11

Frequency sweeps of RGO composite gel compared to a gel without RGO in swollen state

Full size image

The over-proportional decrease and the absolute values of the moduli are related. As can be easily calculated from the theory of linear viscoelasticity and the chemical properties of the material, the NIDO5% gels only show a tiny fraction of the modulus, which theoretically should be based on the amount of functional groups per unit volume. The reason for this lies in the fact that NIDO5% is relatively low in molar mass and, thus, has a significant fraction of chains with no, one, or two DOPA-functionalities per chain, which do not contribute to the network. In the first case, no effect on the modulus should be apparent other than this fraction has to be counted as a “solvent.” In the case of one functionality, the chains can complex with the network, introducing “dead ends,” which reduce the number of rheologically active crosslinkers (Olsen and Johnson 2013; Stadler 2013; Zhou et al. 2012).

Chains with two functionalities lead to chain extension, which does not have a large effect on the modulus. Only the fraction with three or more DOPA groups per chain forms the network, which is small, as statistically, a molar mass M > M w is necessary for three DOPA groups per chain. It is well established that for micellar systems based on, e.g., hydrophobically end-functionalized water soluble polymers, a very low polymer concentration leads to flower-like micelles (Castelletto et al. 2004; Chaterji et al. 2007; Clément et al. 2000; Hamley et al. 2011; Khalatur and Khokhlov 1996; Kujawa et al. 2005; Kumar and Raghavan 2010; Lee et al. 2005; Liao et al. 2007; Suzuki et al. 2012; Tanaka 2000; Tanaka et al. 2011; Tripathi et al. 2006; Tsitsilianis et al. 2000), as the likeliness for intrachain or intramicellar interaction increases with the presence of fewer chains or micelles that do not belong to the same aggregate, i.e., the lower the concentration is.

The same argument can be applied to the effect of swelling, which decreases the effective polymer concentration by a factor of ≈ 3.5. Swelling of the native hydrogel leads to an increase in average interchain distance and, thus, to a decrease in two DOPA groups from different groups interacting. This changes the swelling of supramolecular hydrogels and organogels from that of regular, covalently crosslinked gels, where the crosslinks do not change with time and swelling degree (Hashmi et al. 2012). As the chains themselves are slightly stretched but not completely (otherwise, the stiffness would be determined by the strength of the C-C-bonds of the main chain), the interaction probability of two functional groups belonging to the same chain decreases only slightly. The logical consequence of this is that the increase of the water content decreases the probability of interchain interactions and increases the dilution of the network.

Furthermore, as described earlier, the bonding characteristics change upon addition of even a small amount of water, leading to a reduction in modulus to 78 and 12 % of the anhydrous sample for the native and composite gels, respectively.

When comparing this to the situation on the composite gel reveals the significant difference in the physics governing this sample. Firstly, the water sensitivity of the composite gel is almost negligible in comparison to the native gel (Fig. 7). Therefore, the network consists of two kinds of crosslinks, the regular boron-dicatechol bonds, as in the native gel, and the bonds involving RGO and NIDO5%, as discussed above (Fig. 1b), which unlike the former are not or are only slightly influenced by water. When taking into account the 78 % reduction of boron-dicatechol bonding efficiency (when expressed in terms of modulus), the 12 % reduction in modulus for the composite gel suggests that the majority of the modulus is caused by RGO-related bonds and that these bonds prevent the formation of too many intrachain bonds.

The strain sweep (Fig. 12a) of the swollen native gel shows a first small nonlinearity around γ 0 ≈ 10 %, followed by the main nonlinearity onset around γ 0 ≈ 100 % (Vatankhah-Varnoosfaderani et al. 2013). The low modulus and the low damping of the swollen samples prevented the calculation of a reliable value for G″(γ 0) for γ 0 < 0.7 %.

Fig. 12
figure 12

a Strain sweeps of RGO composite gel compared to a gel without RGO in swollen state and self-healing tests of b swollen native gel and c swollen composite gel

Full size image

Only two such reports of a two-step nonlinearity have been published. Brassinne et al. (2013) found a reduction of modulus by a factor of 10 around γ 0 = 1 % and a nonlinearity limit of around γ 0 = 300 % for a sample consisting of micelles comprised of PS-cores connected by weak hydrophobic interactions and PNIPAM-“hairs” with terminal terpyridine functionalities, complexing with transition metal ions. In this case, the double transition was the result of destructuring of the cores and the hairy coronae’s terpyridine complexes. Vatankhah-Varnoosfaderani et al. (2014a, b) found a two-step transition for NIDO5% in an alkaline DMSO environment with almost the same deformations of γ 0 for the first and second nonlinearity limits.

These results indicate that a two-step transition can occur, which must stem from a partial destructuring of the network at γ 0 ≈ 10 %, followed by a total destructuring involving stretching the sample. As this is significantly different from the situation when the samples are not swollen, the partial swelling-induced stretching must partially explain the behavior as well as the fast kinetics of the sample upon restructuring. This is because a partial restructuring must have occurred between γ 0 = 30 and 270 %, as this is already in the nonlinear range of most samples and, hence, only a very dynamic structure can partially compensate for the loss of structure. The swollen composite gel shows a nonlinearity limit around γ 0 = 300 %, which is significantly higher than the hydrous non-swollen counterpart (γ 0 ≈ 12 %), which is a clear indicator of the different dynamics of these samples. As pointed out earlier, while swelling mostly does not affect the polymer-RGO interactions, it clearly reduces the effect of boron-dicatechol bonds.

As solid fillers usually lead to a reduction of the nonlinearity limit (Anderson and Zukoski 2009; Cassagnau 2003; Choong et al. 2013), the nonlinearity limit of the polymer-RGO interactions is expected to be lower than that for the pure polymeric boron-dicatechol bonds. Careful inspection of the G′(γ 0) data, however, does not reveal any nonlinearity limit in the expected range (γ 0 ≈ 2–50 %). G″(γ 0), however, starts showing some oscillations around γ 0 ≈ 3 %, which indicate that while no effect on the overall modulus can be observed, some processes seem to occur within the sample and do not lower the overall number of crosslinks but increase the damping.

This suggests that some internal reorganization takes place, which could stem, for example, from sliding motions of different GO-NIDO5% aggregates relative to each other. The self-healing properties of the swollen native gel (Fig. 12b) show a quick recovery from an applied deformation γ 0 max of 300 % (within 15 s) with respect to G′, but the recovery of G′ takes significantly longer and does not completely decrease to the pre-shear level, suggesting that the structure of the gel remains slightly disrupted even after a very long recovery. The initial recoveries of G″ and δ are clearly fast, taking place within about 100 s. This ratio of G′- to G″-recovery of about 7 is typical for self-healing systems dominated by elasticity (Guillet et al. 2009). However, unlike many other systems, a very slow decrease by another 10 % is found for t = 800–10,000 s for G″ and δ. In this time range (t = 800–10,000 s), G′(t) decreases slightly (by 3 %). This suggests a two-step recovery, consisting of a fast recovery, in which the structure is rebuilt so that the orientations from the high applied shear are partially frozen in, and a second step, in which the boron-dicatechol bonds are opened, thereby allowing the system to release the orientations step-by-step, which lowers G′(t) slightly and G″(t) more significantly. The data suggest that this process takes significantly longer than the time of the experiment.

This two-step behavior can be interpreted as resembling the common “musical chairs” game, where at the end of the deformation almost every functionality finds a suitable bonding partner almost instantaneously, whereas a few bonding sites remain open due to some fluctuations in arrangement. In contrast to the “musical chairs” game, the functionalities are in a polymer chain and, thus, cannot move freely once functionalities on other parts of the chain have successfully complexed. Hence, matching unmatched functionalities requires opening other complexes and diffusion of the freed chain segments. This renders the process very slow, due to two slow statistical processes having to work together. Figure 12c shows the self-healing test of the swollen composite gel, exhibiting similar characteristics to the native swollen gel.

Within only 30 s, the original value of G′ is reached again. In contrast to the native swollen gel, the recovery of G″ is very fast (≈60 s). However, as in the case of the native swollen gel, G″ and to a smaller extent G′ decrease for a long time. Unlike its unfilled counterpart, however, this long-term decrease leads to a lower G″ (and δ) than before the self-healing experiment. Interpretation of this finding needs to account for the very long “memory” of filled systems, caused by interactions between the filler and the matrix (Münstedt et al. 2008; Schmidt and Münstedt 2002a, b).

Hence, the reduction of G″ and δ can be seen as a consequence of the deformation applied to the sample several hours prior to starting the experiment. In light of this finding, the small minimum in G″(γ 0) and δ(γ 0) between γ 0 = 6 and 30 % (Fig. 12a) could be physical in nature and not an artifact. This minimum could be explained by slightly disturbing the structure of the gel, giving it the chance to reorganize and find an arrangement leading to a lower damping. However, due to the small size of the effect, this can only be speculative.

The thermal- and pH-responsitivities of the resultant composite gel were also investigated, as shown in Fig. 13. The reversibility of the nanocomposite gel proves the pH-sensitivity of the interaction between NIDO5%, boron, and catechol, allowing for several gelation/degelation cycles, as reported previously (a movie is given in SI) (Vatankhah-Varnoosfaderani et al. 2013). In addition, while the organogel does not show any temperature sensitivity, the composite and non-composite gels after swelling in water are thermo-responsive because of the NIPAM and DMA-monomers, which have hydrophobic and hydrophilic parts. Thus, the polymer chains show an LCST in aqueous media and the gel can therefore be switched in to the polymer solution by increasing the temperature to about 35–45 °C. The LCST of gels is affected by water and DMA in the copolymer (Vatankhah-Varnoosfaderani et al. 2013).

Fig. 13
figure 13

Inversion test: (a) nanocomposite gel after adding NaBH4, (b) degelated nanocomposite after adding HCl-gas, (c) degelated native gel after adding HCl-gas, (d) native gel after adding NaBH4, and (e) nanocomposite gel after increasing the temperature to 50 °C

Full size image

Conclusions

The use of RGO in an NIDO5% organogel in DMF leads to relatively similar properties in comparison to those of native gels (without RGO). However, the addition of RGO into the present supramolecular gels affords the advantageous property of significantly weakening the previously established strong decrease of modulus due to the change of bonding characteristic of the boron-dicatechol interactions in the presence of H2O. In general, this weakening of the modulus decrease is enhanced with more water addition to the system.

This suggests that the RGO forms interactions with the NIDO5% copolymer, thereby giving a second method for crosslinking the polymer. The nonlinear behavior further revealed that the RGO induces a clear two-step nonlinearity, which is the consequence of these two types of supramolecular interactions. The tight bonding of NIDO5% to RGO is also evidenced in the TEM micrographs, which show residues of polymer, even after intensive washing.

It can also be concluded, however, that the RGO does not completely dominate the bonding, due to the retention of the self-healing ability and the reversibility of the prepared composite gel.

Based on the rheological data, a rather complicated picture arises of the influence of RGO in the network. As discussed above, RGO plays an active role in the system but does not dominate the whole gel. Hence, it can be concluded that the interaction of the NIDO chains with RGO leads to a transition from a polymer network to a hairy micellar network, which supports the assembly of many chains and, thus, drastically reduces the effective fraction of chains not contributing to the network.

For such a micelle, all chains interacting with one RGO sheet via supramolecular interactions can contribute to the network collectively, i.e., chains with only 1 or 2 DOPA-functionalities, which in solution do not contribute to the stiffness of the sample, contribute to the network when supramolecularly interacting with RGO (making a hairy micelle) in collaboration with other chains (Guillet et al. 2009). Complicated arrangement can explain the differences between the native and RGO gels. In a native gel, classical supramolecular polymer gel dynamics dominate, while part of NIDO5% is attached to RGO, which reduces the effective fraction of network breakers (1 catechol per chain) and chain extenders (2 catechols per chain) due to bonding of these to the surface of the RGO, leading to larger hairy micelles. Based on previous literature on the behavior of hairy micelles with stickers (Brassinne et al. 2013; Guillet et al. 2009), it can be concluded, however, that RGO also partially deactivates the catechols, as catechols on supramolecularly adsorbed chains tend to interact with neighboring adsorbed chains and with the remaining functional groups on the RGO surface (rheological deactivation by intramicellar bonding).

Overall, however, this kind of assembly increases the modulus and, supposedly, the rheologically active catechol interactions, presumably due to the reduction of network breaking and stiffening of the network by RGO, which outweighs the overall reduction of the catechol bonding probability. Adding water leads to a smaller decrease of modulus for the RGO gels than for the native gels. This can be explained by the hairy micelles of the RGO gels compared to the regular supramolecular polymer gels, which, due to the higher number of binding functionalities per unit (micelle vs. single chain), induce a smaller loss of modulus because of the lower fraction of units with few functionalities (2, 3, …). These can turn into network breakers if the bonding probability decreases as a consequence of the equilibrium between boron-dicatechol bonds and boron-mono-catechol-water interactions in the presence of H2O.

At high swelling, the nonlinearity limit of the native gel remains about the same as in the non-swollen, hydrous state, while for the RGO gel, the nonlinearity limit is clearly increased with respect to G′(γ 0) but not to G″(γ 0). This rather contradictory finding can be explained by the additional space for RGO micelles to move under stress, which is not the case when in a tight polymeric network. Hence, the orientations of RGO-NIDO5% micelles show up as oscillations in G″(γ 0), while the overall modulus stays unaffected.

The supramolecular nanocomposite gel presented herein could be potentially useful for the development of artificial scaffolds and drug delivery systems due to the interesting pH responsiveness (Ku et al. 2013). Furthermore, RGO has interesting electrical properties, making this gel suitable for various sensor applications. These potential applications will be investigated in future papers.