The monomer N-isopropylacrylamide (NIPAAm), the crosslinker N,N′-methylenebis(acrylamide) (MBA), the initiator system consisting of ammonium persulfate (APS) and N,N,N′,N′-tetramethylethylenediamine (TEMED), the surfactants Brij®L23 (polyoxyethylene (23) lauryl ether) and AOT (Aerosol® OT; bis(2-ethylhexyl) sulfosuccinate sodium salt), and the nonpolar solvent heptane were bought from Sigma-Aldrich (Munich, Germany) and were used as received for the following syntheses.

Air-dried fragments of the rinsed hydrogel cylinders were put onto adhesive carbon pads, contacted with conductive silver and coated with 3 nm platinum in the sputtering system Leica EM SCD500 (Leica Microsystems GmbH, Wetzlar, Germany). SEM images were recorded in a Neon 40EsB (Carl Zeiss Microscopy GmbH, Jena, Germany) using a Schottky field-emission electron source and an EHT (extra high tension) voltage level of the electron beam of 3 kV. Pore diameters were determined using the ImageJ software package (ImageJ, 2018).

For the swelling experiments, the cylindrical hydrogel samples were cut into disks of 1–2 mm height. The diameter was 5 mm in the swollen state. One hydrogel disk at a time was placed in a copper well filled with water after preheating or precooling the copper well with a Peltier element using a HAT control temperature controller from BelektroniG GmbH (Freital, Germany) and the software BelektroniG HAT Soft Pro v2.3.23. Imaging was carried out using a Bresser Microcam Pro HDMI camera (Bresser GmbH, Rhede, Germany) mounted on a microscope and the respective software Bresser MicroCamLab II. Swelling kinetics were studied by abruptly changing the temperature from 25 to 40 ∘C or vice versa. For the observation of, for example, the deswelling kinetics, a hydrogel disk in a water-filled glass tube was preheated in an oven at 25 ∘C for at least 24 h. Before the measurement, the copper well was preheated to 40 ∘C. Then, the hydrogel disk was taken from the glass tube and put into the copper well while simultaneously the recording of the images started in time lapse (1 picture per 10 s for 2 to 4 h). For the determination of the diameters of the hydrogel disks, we used the self-implemented Python-based software PIA (Python Image Analyzer). The swelling degree of the hydrogel disks was calculated as 1Qd=dtd0, where Qd is the swelling degree determined from the change of the disk's diameter, dt is the diameter of the disk at a certain time of the measurement and d0 is the disk's diameter before the stimulus was applied. Mean values and standard deviations of the swelling degrees were calculated from triple measurements of both the swelling and the deswelling for each hydrogel sample.

Assuming isotropic swelling and deswelling, the swelling degree hypothetically determined from the sample volume (Qv) equals the swelling degree Qd, and hence, we discuss the volume change rather than the diameter change in the following sections.

The time necessary to reach the equilibrium swelling degree (plateau of the curve) was determined graphically for each of the averaged curves.

The PNIPAAm 2 hydrogel was used as a transducer in a piezoresistive setup employing a pressure sensor chip. The piezoresistive sensor setup and the procedure of the sensor measurements are described elsewhere (Franke et al., 2017, Erfkamp et al., 2018).

Figure 1 shows the SEM images of the nonporous PNIPAAm and the porous PNIPAAm 1. As expected, the nonporous hydrogel is clearly not structured, while the porous hydrogel contains an open channel structure with pore diameters of 3.8±2.1 µm. The morphology of PNIPAAm 1 samples has already been presented in Franke et al. (2017). We mention it here in order to compare it with the new PNIPAAm 2.

At first sight, the morphology of the PNIPAAm 2 sample is ambiguous and, hence, we present different views of an air-dried fragment in Fig. 2. The overview of a fracture surface in Fig. 2a shows that the near-surface areas (outer layers) of the sample piece seem dense compared to the structured inner area (inner layer) of the piece. Figure 2b is a close-up of the inner layer of the sample that shows larger, elliptic pores with an average ellipse axis length of 3 µm and smaller structures around these pores. A further close-up of the elliptic pores in Fig. 2c demonstrates their flaked surroundings that show openings with diameters in a range below 100 nm (Fig. 2d). The seemingly dense outer layer of the sample piece shows approximately the same morphology as the inner layer (Fig. 2e), while a view of the sample's surface shows a dense packaging of particles with diameters around 200 nm (Fig. 2f).

We found this morphology for several pieces of the same and different samples. However, different morphologies at different parts of the SEM samples were not expected because the sample pieces were cut out of disks as described above, and we expect them to be homogeneous in their inner structure. We assume that the air-drying process altered the original morphology of the wet samples. Near-surface parts collapsed more than the inner part of the samples where the original morphology is presumably conserved, while the morphology of the surface of the pieces is completely collapsed forming a dense, particle-like structure. Hence, the imaging of the morphology of PNIPAAm 2 has to be the subject of further studies. Therefore, a faster drying method, preferably freeze-drying using liquid ethane, is necessary in order to preserve the inner structure of the hydrogel sample pieces. At this point, we consider the morphology shown in Fig. 2d to be the original one.

Comparison of the morphology of PNIPAAm 1 (Fig. 1b) and PNIPAAm 2 (Fig. 2d) shows that the pore diameters of PNIPAAm 1 are 1 order of magnitude larger than the pore openings in the flaked structure of PNIPAAm 2. It seems that the additional use of an organic solvent stabilized the template structure as predicted, and the phase separation of the template solution during polymerization is at least partially suppressed. The occurrence of the larger elliptical pores, however, indicates that a partial phase separation occurs because micelles of a size in the micrometer range are thermodynamically not expected in the template solution. Small pore sizes in the 100 nm range in the flaked part of the morphology indicate that at least parts of the template structure were preserved.

Phase separation in microemulsions and lyotropic liquid phases during synthesis is a common constraint of these surfactant-based template systems (Hentze and Antonietti, 2001; Sjöblom et al., 1996; Pavel, 2004). Further research is required regarding the possibilities of preventing phase separation of the template structure in order to preserve it in the morphology of the porous hydrogel.

Figure 3 summarizes the results of the swelling experiments (deswelling at 40 ∘C and swelling at 25 ∘C) and compares the swelling kinetics of nonporous PNIPAAm and porous PNIPAAm 1 and 2.

Figure 3a shows the deswelling behavior of the nonporous and the porous PNIPAAm samples. The nonporous PNIPAAm sample shows the two-step swelling behavior that we found before in Franke and Gerlach (2020) with an initial fast decrease of the swelling degree Qd followed by a slow further deswelling caused by the so-called skin effect. In contrast, the porous PNIPAAm samples deswell fast and in a one-step procedure. During the course of the measurements, which took up to 4 h, the nonporous PNIPAAm never reaches the equilibrium swelling degree (plateau).

Porous PNIPAAm 1 and 2 reach the equilibrium swelling degree after 6 and 2 min, respectively, and therefore much faster than the nonporous sample (Fig. 3c). For the swelling similar results occurred (Fig. 3b and d). While the swelling of the nonporous PNIPAAm does not reach an equilibrium during the measurement period (Fig. 3b), the porous PNIPAAm 1 and 2 reach the equilibrium swelling degree after 6 and 2 min, respectively (Fig. 3d).

Additionally, it is obvious that the volume of the nonporous hydrogel sample is reduced much more during deswelling than the volumes of the porous PNIPAAm samples (Fig. 3a) and vice versa (Fig. 3c). We explain this finding as follows. The porous PNIPAAm samples contain less deswellable/swellable hydrogel material per volume unit. Therefore, the volume of the samples is decreased/increased less. We also have evidence from our previous and ongoing research that the pore structure does not collapse during the deswelling and, therefore, does not contribute to the deswelling and swelling process. On the other hand, the entire volume of the nonporous sample is capable of deswelling/swelling. Assuming that the equilibrium swelling degree of the deswelling is reached after 54 h at 40 ∘C (sample kept in the oven for this period), the volume of the nonporous PNIPAAm is reduced to 50 % compared to its volume at 25 ∘C. The volumes of the porous samples change around 13 % with the exception of the swelling of PNIPAAm 2 where the volume increases around 17 % after decreasing only about 13 % during deswelling. Incomplete conditioning of the samples might cause this difference. Another problem is the monitoring of the swelling/deswelling of a fast-swelling/fast-deswelling sample that has to be put into the measurement setup from a container of different temperature. Since the swelling/deswelling takes only 2 min and the positioning of the samples took between 10 and 30 s, deviations might occur during evaluation.

For a sensor application, this reduced increase and decrease of the swelling degree is an advantage regarding piezoresistive pressure sensor chips that have a limited capacity for withstanding the volume change of the transducer.

The explanation for the fast swelling and deswelling of porous PNIPAAm 1 and 2 is presented in a schematic illustration of hydrogel volume units (presented as area units) in Fig. 4 for the release of water from the hydrogel sample (deswelling). Compared to the nonporous hydrogel in Fig. 4a, the porous hydrogel 1 in Fig. 4b consists of an additional inner surface area due to the pores and, hence, more water can be released over the overall surface (outer and inner surface forming the specific surface area expressed in m2g-1) per time unit. The channel-like structure of the open-pore morphology contributes to the faster transport of water to the outside of the hydrogel volume. The reduction of the pore size is accompanied with an addition of swellable material per volume unit and leads to an increase of the inner surface area (Fig. 4c). As a result, even more water can be released per time unit. The transport of water is again supported by the channel structure of the pores.

In summary, we assume and propose that an increase of the specific surface area of a hydrogel sample is the key factor for decreasing the swelling and deswelling time. The incorporation of pores as channel structures into the hydrogel volume additionally supports the transport of water to the inside or outside of the hydrogel volume. Small dimensions of the swellable material forming the pores in the sample volume prevent the skin effect due to short paths of diffusion.

Nonetheless, we are aware of the possibility of competing mechanisms coming from fluid mechanics (especially Darcy's law). We presume that there is an optimum (minimum) of the swelling and deswelling time for a certain specific surface area/pore size. For pore sizes below that, the swelling and deswelling times may rise again due to Darcy's law. Up to now, we cannot predict that key pore size, but the surfactant-based template method to our knowledge is the only synthesis method producing small pore sizes in the nanometer range and a large variety of pore geometries required for further research.

Figure 5 shows the results of the sensor measurements for two sensors containing PNIPAAm 2. For both sensors the sensor signal changes with changing temperature caused by swelling (at 25 ∘C) and deswelling (at 50 ∘C) of the hydrogel. A plateau of the sensor signal is reached for each temperature step, and the level of the sensor signal stays the same after repeated swelling and deswelling processes. However, there is an obvious difference between the signal of sensor 1 and sensor 2. ΔV, the difference between the output voltage at the swollen state and the output voltage at the deswollen state, is different for sensor 1 and sensor 2. Since the signal of sensor 1 is less smooth than the signal of sensor 2, showing a pronounced overshooting at 25 ∘C and recurring interference of the signal at 50 ∘C, we ascribe this difference to an inaccurate assembling of sensor 1. The pressure sensor chip may be tilted, and hence the hydrogel cannot implement the full range of swelling on the pressure sensor chip leading to a smaller sensitivity of sensor 1 and interference during swelling and deswelling.

As shown in Table 1, the response times for the deswelling of PNIPAAm 1 and 2 at 50 ∘C are similar. While PNIPAAm 1 deswells in less than 20 s PNIPAAm 2 is a little slower with around 40 s. The response time for the swelling at 25 ∘C, however, is significantly decreased with 1.5 to 4 min when using PNIPAAm 2 compared to PNIPAAm 1, which needed 21.5 min. These results show that the incorporation of pores in the nanometer range into a hydrogel leads to a major improvement of the performance in a sensor application compared to macroporous and nonporous hydrogels.

Inside the pressure sensor chip, mechanical stability of the hydrogel is a key characteristic beside the capability of fast swelling and deswelling. PNIPAAm 1 lacked this mechanical stability leading to long-term swelling. As explained in Franke et al. (2017), the porous hydrogel has to work against the flexure plate of the pressure sensor chip, which compresses the pores during deswelling. We showed that a full recovery of the porous structure is possible but takes a long time compared to the deswelling, which is supported by the flexure plate.

Figure 4c shows that a hydrogel with smaller pores contains additional hydrogel material per volume unit compared to the hydrogel with larger pores in Fig. 4b. This additional material serves as pillars stabilizing the porous hydrogel structure when working against the flexure plate. As a result, PNIPAAm 2 swells much faster in the sensor using both the increased mechanical stability due to the additional hydrogel material per volume unit and the increased swelling kinetics due to porosity.

The response time for the deswelling of PNIPAAm 2 is slightly increased compared to PNIPAAm 1, which can also be explained with the additional hydrogel material per volume unit. The flexure plate of the pressure sensor chip helps press the water out of the hydrogel. Due to lower mechanical stability because of larger pores, this effect is more pronounced for PNIPAAm 1 than for PNIPAAm 2. PNIPAAm 2 contains more hydrogel material per volume unit, which puts a force against the flexure plate. Therefore, the supporting role of the flexure plate during deswelling is less pronounced, and the response time for the deswelling increases compared to PNIPAAm 1.