Surface-controlled spatially heterogeneous physical properties of a supramolecular gel with homogeneous chemical composition

Controlling supramolecular self-assembly across multiple length scales to prepare gels with localised properties is challenging. Most strategies concentrate on fabricating gels with heterogeneous components, where localised properties are generated by the stimuli-responsive component. Here, as an alternative approach, we use a spiropyran-modified surface that can be patterned with light. We show that light-induced differences in surface chemistry can direct the bulk assembly of a low molecular weight gelator, 2-NapAV, meaning that mechanical gel properties can be controlled by the surface on which the gel is grown. Using grazing incidence X-ray diffraction and grazing incidence small angle X-ray scattering, we demonstrate that the origin of the different gel properties relates to differences in the architectures of the gels. This provides a new method to prepare a single domain (i.e., chemically homogeneous) hydrogel with locally controlled (i.e., mechanically heterogeneous) properties.


Preparation of spiropyran-modified surfaces
To create spiropyran-modified surfaces, clean silicon wafers were modified with amine functionalities through which (2) was immobilised on the surface ( Figure S4). The samples were placed in petri dishes and cleaned by sequentially sonicating them for five minutes in methanol, acetone and isopropanol. The substrates were dried using compressed air. Sample cleaning and exposure of hydroxyl groups on the surfaces was accomplished via piranha cleaning. Piranha solution (20 mL) (Caution: Piranha solutions reacts violently with organic materials) was added to a glass petri dish containing the samples and left to react for 40 minutes. The samples were rinsed and sonicated in Millipore ultrapure water for ten minutes, then dried using compressed air. Figure S4. Substrate functionalisation with SP-COOH (2) via 3aminopropyl)trimethoxysilane (APTMS) followed by diisopropylcarbodiimide (DIC) coupling of (2) onto the surface. The resulting SP-COOH modified surface (SP) was converted to the merocyanine form (MC) upon exposure to 365 nm light. Below the schematic, representative water droplets and measured water contact angles on the surfaces show the change in hydrophobicity of the surface at each modification step (n = 5, 6, 6, and 4 for the silicon wafer, APTMS, SP and MC surfaces, respectively).
Amine functionalisation of the samples was accomplished via silanisation of the slides with (3-aminopropyl)trimethoxysilane (APTMS). The slides were placed in a glass petri dish and a 1% solution of APTMS in toluene (20 mL) was added. The samples were incubated at room temperature for one hour, followed by washing and sonicating sequentially with toluene, acetone and Millipore ultrapure water for five minutes each, and drying under compressed air.
The final step of the modification involved the addition of SP-COOH to APTMS modified surfaces. A solution of SP-COOH in anhydrous DMF (95 mM, 13.9 mL) and diisopropylcarbodiimide (DIC) (410 µL) were added to a glass petri dish containing the APTMS functionalised slides. This was left at room temperature on a Heidolph Rotamax 120 Orbital Shaker at 20 rpm for 16 hours. After incubation, the slides were sequentially washed in DMF, acetone and water for five minutes each, and then dried using compressed air. The slides were stored in the dark in a desiccator under vacuum.

Water contact angle
At each step of the surface modification, water contact angle (WCA) measurements were used to monitor changes in surface properties. The WCA was determined with a KSV Cam200 Optical Contact Angle Meter. The Cam200 was set up to record 20 frames at a speed of one frame per second for each droplet. The WCA of the droplet in each frame was calculated using the circle fitting method. As the droplets are not immediately in equilibrium after deposition on the surface, values from the first two frames were discarded to reduce the likelihood of the data to be affected by WCAs measure in a non-equilibrated state. The WCA for the remaining frames were used to extrapolate the WCA at time-point zero, which was used as the WCA value for one measurement. The reported WCA measurements are presented as average ± standard deviation (SD). n = 5, 6, 6, and 4 for the silicon wafer, APTMS, SP and MC surfaces, respectively.
The WCA angles increased at each modification step from 4.7° ± 0.7°, to 60.8° ± 4.8° and 90.2° ± 5.4° for silicon, APTMS and SP, respectively ( Figure S4). The increasing hydrophobicity indicates changes to the surface chemistry at each modification step.

ToF-SIMS
To analyse the chemical composition of the surfaces, ToF-SIMS was carried out with a ToF-SIMS IV instrument (ION-TOF GmbH, Münster, Germany) using a 25 keV Bi3 + primary ion source with a target current of approximately 1.0 pA. Charge compensation was applied via a low energy (20 eV) electron floodgun. SIMS data were acquired in negative polarity over regions of 500 µm × 500 µm at 256 × 256 pixel resolution for 20 scans or over large areas of 2 mm x 6 mm (or 1 mm x 6 mm) at 100 pixels / mm for1 scan. The data was analysed with SurfaceLab 6 (ION-TOF GmbH). Negative ion mass spectra were calibrated to CH -(m/z 13), CH2 -(m/z 14), C3H -(m/z 37) and C4H -(m/z 49).  Figure S7 and the corresponding signals in the ToF-SIMS spectra are shown in more detail in Figure S8.
The modification of silicon with APTMS showed an increase in CNion intensity and a decrease in SiO2ion intensity, confirming that the surface was modified with a nitrogen containing compound (APTMS). The appearance of the molecular ion of the full spiropyran molecule (C18H15N2O3 -) on the SP surface indicates attachment of SP-COOH on the surface. CNion intensity remains at similar levels because CNis a generic indicator of nitrogen containing material on the surface and SP-COOH also contains nitrogen. Figure S5. Surface analysis of the modified surfaces by ToF-SIMS. Negative ion spectra in the lower m/z region for silicon, the APTMS modified surface (APTMS), the spiropyran surface (SP) and the SP surface after exposure of UV light to convert spiropyran to the merocyanine form (MC). Marker fragments used to distinguish between samples are highlighted. Figure S6. Surface analysis of the modified surfaces by ToF-SIMS. Negative ion spectra in the higher m/z region for silicon, the APTMS modified surface (APTMS), the spiropyran surface (SP) and the SP surface after exposure to UV light to convert spiropyran to the merocyanine form (MC). Marker fragments used to distinguish between samples are highlighted.

Photo-induced surface switching
SP surfaces were irradiated with UV light (365 nm, 8 W) or white light (300 -600 nm, 3 W) for 12 hours to drive the population of molecules on the surface to predominantly adapt either the spiropyran (SP) or merocyanine (MC) form.
After conditioning of the SP surface under UV light, the resulting MC surfaces showed a significant decrease in WCA ( Figure S4). In the ToF-SIMS spectra the presence of the MC form was characterised by the appearance of ions associated with the chromene moiety of the molecule (C6H3NO3 -, C7H4NO3 -) that were not observed at high intensities on the SP surface ( Figures S5, S6 and S8). The ion indicative for the whole SP-COOH molecule (C18H15N2O3 -) was present on both the SP and MC surfaces. This indicates that while the ion intensity of the molecular ion is not affected by the switching, the generation of smaller fragments is facilitated by the conversion of the ring structure into an open structure on the MC surface.

Morphological and mechanical properties of dry gel films
AFM experiments were performed on a Bruker Dimension ICON with PeakForce QNM module installed. An AFM probe (Bruker RTESPA-150) with a spring constant of 6.77 N/m and a typical tip radius of 8 nm was used for the imaging of morphology and the measurement of mechanical properties in this study ( Figure S9). The images in Figure S9 were processed in Gwydion 2.59; the polynomial background was removed (horizontal & vertical degree of 2), rows were aligned (method: matching) and horizontal scars were corrected before applying a filter (sharpen, size: 5 pixels). The gel was prepared as described in the main document on either the SP or MC surfaces and dried under ambient environment.
The deflection sensitivity of the tip was first calibrated on a clean silicon wafer. Then, various indentation depths of 20-400 nm were tested to rule out the substrate effects on the gel film mechanical properties. For the actual measurement, seven different locations across the whole sample were chosen to obtain the Young's modulus using the PeakForce QNM method. On each of the locations, the image was collected on a 2 µm × 2 µm area at a scan rate of 1.0 Hz. The resolution of the image was 512 × 512 pixels. 700 evenly distributed locations were chosen to perform an independent indentation test. Various indentation depths of 20 -400 nm were tested to rule out any potential effects of the silicon wafer substrate on the measurement and a 200 nm indentation was chosen for the measurement.
In this mode, the AFM cantilever was controlled to oscillate at a certain frequency and perform a tapping mode scan on the sample surface. During each tapping cycle, the force and separation curves were recorded. The surface modulus at the indentation point was derived from the unloading curve by using the Derjaguin-Mueller-Toporov (DMT) fitting model: 3, 4 where F is the force on the tip, R is the tip radius, d is tip deflection, and ℎ is the adhesion force between tip and sample. In this equation, these variables are recorded during experiments, leaving the reduced modulus, , as the only unknown variable, which can be calculated by fitting the unloading curve using a power function in NanoScope Analysis 1.7. With the knowledge of the reduced modulus ( ), the tip modulus ( ) and Poisson's ratio (), the Young's modulus of the sample ( ) could be obtained after rearranging the following equation:

Quasi-static Nanoindentation Experiments
Quasi-static nanoindentation experiments were performed using a nanoindentation device (Chiaro, Optics 11) mounted on top of an inverted phase contrast microscope (Evos XL Core, Thermofisher). Measurements were performed at room temperature in air on hydrated hydrogel samples. For each sample, a set of single forcedisplacement (F-z) curves was acquired at a speed of 2 µm/s over a vertical range of 10 µm, changing the x point of every indentation by ∼ 250 µm. The cantilever selected for the experiments had an elastic constant k of 0.34 N/m and held a spherical tip with a radius (R) of 22 µm, so that the mechanical properties of the samples are averaged over a reasonable area ( = √2 , where a is the contact radius and is the indentation) (see Figure S10 a and b).
The forward segment of the collected curves was analysed using a custom software 5 programmed in Python 3 and the Numpy 6 /Scipy 7 scientific computing stack. Briefly, data were first aligned using a baseline detection method based on the histogram of the force signal. F-z curves were then filtered using a Savitzky Golay filter with a window length of 100 nm to remove random noise. After, F-z curves were converted to force-indentation (F-) curves by finding the contact points (z0, F0) using a thresholding algorithm, and computing the indentation according to: To quantify the elastic properties of the gels, F-curves were then fitted with the Hertz model (Equation 4) up to an indentation of about 10% of the tip radius, and the Young's Modulus (E) extracted. The Poisson's ratio ( ) was taken as 0.5 assuming material's incompressibility given the highly hydrated nature of the gels.

Dynamic Nanoindentation Experiments
Dynamic nanoindentation experiments were performed using a nanoindentation device (Chiaro, Optics 11) mounted on top of an inverted phase contrast microscope (Evos XL Core, Thermofisher). Measurements were performed at room temperature in air on hydrated hydrogel samples. For each sample, a set of curves was acquired by performing two matrix scans in different regions of the gel. Each matrix scan consisted of 25 indentations, with a lateral step in x and y of 50 µm. Indentations consisted of stress relaxation experiments in dynamic mode, where a step indentation of 2000 µm with superimposed sinusoidal oscillations of amplitude 5 nm at 1,2,4 and 10 Hz was used as the input signal. The relaxing force was recorded over the time of the indentation, which was set to approximately 20 s (see Figure S10c).
The cantilever selected for the experiments had an elastic constant k of 0.34 N/m and held a spherical tip of radius (R) 22 µm in radius, so that the mechanical properties of the samples are averaged over a reasonable area ( = √2 , where a is the contact radius) (see Figure S10a).
Data were analysed using the manufacturer's software (DataViewer V2.4, Optics 11). In brief, the sinusoidal oscillations were fit to a cosine function, and the estimated amplitude and phases were used to obtain the storage (E I ) and loss (E II ) moduli, which are a measure of the material's elasticity and viscosity, respectively: Here, f is the frequency of oscillation; F0 and 0 are the amplitudes of the oscillatory force and indentation signal, respectively; is the phase shift between the imposed oscillatory indentation signal and recorded force; is the Poisson's ratio, taken as 0.5 as previously described; and a is the contact radius, as previously defined. Since the contact area changes with the oscillatory signal, the contact radius is calculated using the average indentation depth given the small oscillatory amplitude of the indentation (5 nm). 8 G I and G II were obtained following E=3G, and all fits with an R 2 lower than 0.8 were rejected. The shear moduli at a frequency of 4 Hz were used for statistical comparison (see Figure 2 in the main text).

Preparation of patterned surfaces
Spatially defined conversion of an SP-COOH modified sample into either the SP or the MC surface was performed by selective irradiation of the two different areas on the same sample. Irradiation was carried out for at least 12 hours and performed simultaneously by controlling exposure to white (300 -600 nm, 3 W) and UV (365 nm, 8 W, 0.458 mW cm -2 at sample position) light on the surface via optically sealed containers ( Figure S11).
Gels were prepared on the surfaces immediately after patterning. 2-NapAV (2.5 mg) was suspended in deionized water (0.5 mL). An equimolar quantity of NaOH was added and the mixture was gently stirred for two hours until a clear solution was formed. The pH of this solution was 10.7. To prepare hydrogels, 2.32 mg glucono-δ-lactone (GdL) powder was added to tune the final pH to 4. The mixed 2-NapAV-GdL solution (350 µL) was then applied on a modified surface and left overnight in a sealed petridish to ensure complete gelation before the nanoindentation measurements. Figure S11. Schematic side view of the preparation of patterned samples. An SP-COOH modified silicon wafer was simultaneously exposed to white (300 -600 nm, 3 W) and UV (365 nm, 8 W, 0.458 mW cm -2 at sample position) light for 12 hours. Selective irradiation of the surface was achieved by containing the radiation within optically sealed containers.

Chemical characterisation of surfaces by ToF-SIMS
The patterned sample (without gel) and the gel prepared on the patterned sample were analysed by ToF-SIMS to determine if the materials were chemically different on the SP and MC sides of the sample. The sample acquisition parameters were the same as those detailed above. The analysis area used (6 mm × 2 mm raster scans with 150 pixels × 50 pixels for SP-COOH surfaces and 6 mm x 1 mm with 150 pixels x 25 pixels for gel surfaces) spanned across the interfacial region of the SP and MC parts of the surface. For the patterned surface without gel, the same ions that were identified above were used to create ion maps that visualise the distribution of SP-  For semi-quantitative analysis of the intensities of the selected ions, four regions of interest of identical size (400 µm × 400 µm) were created on the SP and MC side. The ion intensities for these four regions were used to calculate an average ion intensity for the 3 ions of interest on both sides of the surface and are shown in Figure S12.
Only the C7H4NO3ion showed a statistically significant difference on the MC and SP sides, whereas the other two ions are similar on both regions. This indicates that the whole silicon wafer sample was covered with SP-COOH and confirms that after patterning, the UV exposed part of the surface (MC side) indeed changed towards the open form. The lack of a difference in the C10H7Oion intensity on the dry gel samples indicates that the gel formed on the patterned surface is chemically homogenous despite the chemical differences on the patterned surface at the start of the gel formation process.

Mechanical characterisation of patterned hydrated gels by oscillatory nanoindentation
The mechanical properties of wet gel films on the patterned samples were measured by oscillatory nanoindentation. Tests were performed to determine the complex shear modulus of these gels using a KLA-Tencor Nanoindenter G200 with a DCM-II Head (CA, US). Measurements were conducted with a 100 µm flat punch indenter (Synton-MDP Ltd., Nidau, Switzerland) at room temperature (25 o C) at 110 Hz with a precompression of 5 µm. 30 indentations across the centre of the sample were made from the edge of one surface to the other, covering a distance of about 10.8 mm. The distance between each indent is 400 µm. The storage modulus (G'), the loss modulus (G'') and the loss factor tan(δ) (i.e., the ratio of G''/G') were calculated for each indentation. After each indent, the tip was cleaned by indenting a piece of double-sided Scotch tape mounted on an adjacent sample puck before returning to the gel sample. Figure S13a shows the layout of the samples used for the nanoindentation measurements. The G' and G'' values measured across the patterned sample are shown in Figure S13b and S13c.
Control measurements were performed in the same fashion on gel prepared on individual, non-patterned SP and MC surfaces. To do this, the full SP-COOH sample was exposed to either UV or visible light before gel formation to obtain a chemically uniform SP or MC surface. Nanoindentation measurements were performed across the samples as described for the gels on patterned samples; the results are shown in Figure S14. Mean tan(δ) values were determined (0.173 ± 0.040 and 0.252 ± 0.020 for SP and MC surfaces, respectively) and compared via a two sample t-test. The values were found to be statistically significantly different (p = 7.14 x 10 -8 ) between the gels on the SP and MC surfaces. In addition, to rule out potential measurement artefacts, the data from the control samples was divided into two groups of equal size at the centre of the measured length range. When comparing the mean tan(δ) values from these two groups, no statistically significant difference was found (two sample ttest, p = 0.097 and 0.121 for SP and MC, respectively). This confirms that the difference between the SP and the MC sides on the patterned sample is real and not a measurement artefact. Figure S14. Loss factors (tan(δ)) obtained from gels formed on uniform SP and MC surfaces that were not patterned before gel formation.

Gel fibre structure
To evaluate the structure of the gel samples, grazing-incidence X-ray diffraction (GID) and grazing-incidence small-angle X-ray scattering (GISAXS) measurements were performed on gels formed on individual SP and MC surfaces. Details of the experimental procedure are provided in the main manuscript.
The sample was prepared using the method described in Section 5.1 on a SP-COOH treated silicon wafer. GID and GISAXS experiments were performed on the I07 beamline, Diamond Light Source, Didcot, UK. Time-resolved GID experiments were started immediately after the spreading of the hydrogel forming solution using X-rays with an energy of 18 keV, wavelength of 0.6888 Å to achieve a q range of 0.05 Å -1 -5.8 Å -1 . For time resolved GISAXS experiments, patterns with a q range of 0.027-0.6 Å -1 were obtained using X-ray energy of 14.5 keV and a wavelength of 0.8551 Å. The sample-detector distance was 3 m.

Grazing-incidence X-ray diffraction (GID)
Grazing-incidence X-ray diffraction (GID) is used to probe the structure of the gel both parallel with the surface (along qparallel) or perpendicular to the surface (along qperpendicular). The diffraction patterns obtained for the GID experiments on gels on the SP and MC surfaces are shown in Figure S15. The images display some directional patterns. On the SP surface, the reflections at 0.55 Å -1 and 1.34 Å -1 are more dominant in the qparallel direction. On the MC surface, these two reflections are present in both the parallel and the perpendicular directions. On both surfaces, the position of these two peaks is slightly shifted in the patterns for the qperpendicular and qparallel directions compared to the patterns obtained when integrating over the whole diffraction image (Figure 4c and 4f). This is attributed to a small degree of orientation of the fibres that may have been introduced during the drying process. For the structural determination, the average values from the patterns obtained after full integration were used.
On the MC surface, the fully integrated spectra show new reflections at 0.09 Å -1 , 0.18 Å -1 and 0.27 Å -1 (Q ratio of 1:2:3) (Figure 4f). These reflections are only present in qperpendicular, indicating that these peaks correspond to a directional order of the gel fibre bundles. Figure S15. (a) 2D Grazing-incidence X-ray diffraction pattern (GID) of the 2-NapAV hydrogels formed on the SP and MC surfaces at 103 min and 76 min, respectively. The 2D images correspond to the patterns shown in Figure 4c. (b) Diffraction patterns obtained by integrating the 2D images in the parallel (blue) and perpendicular (red) direction. Annotations correspond to peak positions observed in the full integration of the 2D images and are intended for comparison with peak positions present in the parallel and perpendicular directions.

Grazing-incidence small-angle X-ray scattering (GISAXS)
While not strictly applicable to the grazing-incidence geometry, we can qualitatively analyse the small-angle X-ray scattering (SAXS) part of the data using standard bulk SAXS models. This is a useful indicative approximation of the full quantitative analysis. Thus, the data was fitted using a Kratky-Porod flexible cylinder model with polydisperse cross section and a uniform scattering length. 9, 10 This model describes a worm-like flexible chain of Length (L, the total length), made from freely jointed cylindrical units with a stiff segment length (lp), which is half of the Kuhn Length (b), and a cross-sectional radius of R ( Figure S16). Polydispersity of the cross-section is included using a Schulz distribution. An overall scale factor that corresponds to the volume fraction of the cylinders is also included in the fitting parameters. For the fitting process, the restraining conditions were that the Kuhn length had to be larger than 2R, and smaller than the overall length L. The scattering length densities (SLD) used in the fitting were 10.470 × 10 -6 Å -2 and 9.469 × 10 -6 Å -2 for 2-NapAV and water, respectively. The fitting results are shown in Figure S17 and Table S1.

Replicates and statistical analysis
Where averages are reported, all values are presented as mean ± standard deviation.
Water contact angle measurements were performed on two separate samples with a total number of repeat measurements of n = 5, 6, 6, and 4 for the silicon wafer, APTMS, SP and MC surfaces, respectively.
Peak force measurements were carried out on seven different locations on the SP and the MC surfaces. 700 measurements were performed at each location. All data from one surface was pooled and used to plot a frequency distribution of the occurrence of Young's moduli on each of the two surfaces.
ToF-SIMS data from non-patterned samples was collected from a 500 μm x 500 μm area with 256 × 256 pixels each to qualitatively confirm surface modification. Each pixel represents an individual mass spectrum. N = 1, 2, 3 and 3 for silicon, APTMS, SP and MC surfaces, respectively. One representative image has been used to generate the spectra shown. For patterned surfaces, large area scans with a length of 6 -8 mm across the interface and a height of 1.2 -3.2 mm were acquired from three separate samples each for the SP-COOH and the gel surface. The dimensions varied to provide a sufficiently large area to ensure measurement of the interface and at the same time minimise the analysis time required. From these images, smaller areas of 6 mm x 2 mm (1200 pixels x 400 pixels) and 6 mm x 1 mm (1200 pixels x 200 pixels) were cropped from the SP-COOH and the gel surface samples, respectively. The pixels were binned to combine 64 measurements (i.e. 64 individual pixels) into one larger pixel, giving an image resolution of 150 pixels x 50 pixels and 150 pixels x 25 pixels for the SP-COOH and the gel surface, respectively. While this reduces the spatial resolution of the image, it increases the signal intensity for each resulting pixel and allows a clearer visualisation of the change in ion intensities on the surface. The data obtained from one of the three samples was selected to generate representative spectra and images and to perform statistical analysis of ion intensities on the two sides of the patterned surfaces. Four regions of interest were created on each side of the patterned surface as described in section 5.2 and their intensities were used to report a mean ion intensity ± standard deviation (n = 4). For statistical comparison of the ion intensities on the two sides of the patterned surface a 2 sample t-test was used with a significance level of 0.05 and assuming equal variance. Nanoindentation on patterned surfaces was performed on three separate samples; one representative sample was used to generate the figures and perform a statistical comparison between the two sides of the patterned surface. n = 14 on the SP and n = 12 on the MC surface as two data points in the transition area were excluded for the statistical analysis. Non-patterned control samples were measured once each with 16 measurements on each sample. For statistical comparison, the data was either pooled into two groups of 8, separating the data points at the centre of the measured line, to compare two different sections of the control samples (DF = 14; p-value = 0.097 and 0.121 for the SP and MC surface, respectively). To compare the values between the SP and MC control samples, all 18 data points from each sample were used. In all cases, statistics was performed with a 2 sample t-test using a significance level of 0.05 and assuming equal variance. Degrees of freedom (DF) and p-values of significantly different datasets are reported in the caption of the corresponding figures.
GID and GISAXS data were collected from one sample each for the SP and MC surfaces.