Pathway complexity in fibre assembly: from liquid crystals to hyper-helical gelmorphs

Pathway complexity results in unique materials from the same components according to the assembly conditions. Here a chiral acyl-semicarbazide gelator forms three different gels of contrasting fibre morphology (termed ‘gelmorphs’) as well as lyotropic liquid crystalline droplets depending on the assembly pathway. The gels have morphologies that are either hyperhelical (HH-Gel), tape-fibre (TF-Gel) or thin fibril derived from the liquid crystalline phase (LC-Gels) and exhibit very different rheological properties. The gelator exists as three slowly interconverting conformers in solution. All three gels are comprised of an unsymmetrical, intramolecular hydrogen bonded conformer. The kinetics show that formation of the remarkable HH-Gel is cooperative and is postulated to involve association of the growing fibril with a non-gelling conformer. This single molecule dynamic conformational library shows how very different materials with different morphology and hence very contrasting materials properties can arise from pathway complexity as a result of emergent interactions during the assembly process.

To a of solution of malonic dihydrazide (0.25 g, 1.9 mmol) in dimethylacetamide (DMA) (2 mL), 2.1 equivalents of (S)-(-)-α-methylbenzyl isocyanate or (R)-(+)-α-methylbenzyl isocyanate (for the synthesis of 1 SS or 1 RR respectively) were added (0.565 mL, 4.0 mmol) and the mixture was stirred at room temperature for 24 hours.Water was added (100 mL) until a heavy white precipitate appeared.The precipitate was filtered and washed with three portions of water (3 x 50 mL).The solid product was then purified via flash column over silica gel using a mixture 95:5 acetonitrile:water as mobile phase.The purified product was then recovered and redissolved in the minimum amount of ethanol (20 mL).Water was added until the solution turned translucent (250 mL approx.)and then left to precipitate slowly overnight.After vacuum filtration, the final product was obtained as a white powder (0.690 g, 1.6 mmol, 85%).To a of solution of malonic dihydrazide (0.25 g, 1.9 mmol) in dimethylacetamide (DMA) (2 mL), 1 equivalent of (S)-(-)-α-methylbenzyl isocyanate was added (0.267 mL, 1.9 mmol) and the mixture was stirred at room temperature for 24 hours.The solvent was removed in vacuo and the residue was purified via flash column over silica gel using a mixtures of acetonitrile:water as mobile phase (from 100% acetonitrile to 95:5).The purified intermediate was dissolved in dimethylacetamide (10 mL) and (R)-(+)-α-methylbenzyl isocyanate (0.267 mL, 1.9 mmol) was added to the solution at room temperature, before stirring for 24 hours.Water (200 mL) was added to precipitate the product, which was isolated by vacuum filtration and washed with water (300 mL).The precipitate was purified by column chromatography, eluting dichloromethane:methanol mixtures (from 100% dichloromethane to 95:5) .The purified compound was dissolved in diethyl ether and the solvent was removed in vacuo to obtain 2,2'-malonylbis(N-((R,S)-1phenylethyl)hydrazine-1-carboxamide) as a white powder (0.45 mmol, 12% yield).

Crystallization.
Structure of 1 RR : 5 mg of 1 RR was dissolved in 1 mL ethanol.Water was added in 50 μL aliquots until the solution turned translucent (0.8 mL total).The solution was then heated at 75 ºC until turned transparent, before being cooled slowly from 75 ºC to 50 ºC over 10 days.This yielded crystalline needles which were identified to be an ethanol solvate of 1 RR .
Structure of 1 RR -1 SS racemic mixture acetonitrile solvate (stable conformer A and metastable conformer B): 1.5 mg of 1:1 1 SS and 1 RR was dissolved in 2 mL with heat and slowly cooled from 80 ºC to 50 ºC over 5 days during which time a concomitant mixture of needle and plate crystals formed.Crystals were stored at room temperature in the mother liquor thereafter.The needle crystals were a stable acetonitrile solvate of 1 RR -1 SS racemic mixture in conformation A while the plates were unstable and decomposed in the synchrotron beam.The structure obtained is an acetonitrile solvate of 1 RR -1 SS racemic mixture in conformation B.
Structure of 1 RR dihydrate: 1 RR (10 mg) was dissolved in 3 mL of a 1:1 1,4-dioxane/water mixture using heat.The solution allowed to cool to ambient temperature and left undisturbed for 5 days, after which crystalline needles had formed.
Structure of 1 RR ethanolate: 1 RR (5 mg) was dissolved in 1 mL EtOH and 3 × 0.5 mL aliquots of water (1.5 mL total) was added to the vial.The resulting translucent solution was then heated until transparent and allowed to cool to room temperature.The vial was left undisturbed for 3 months after which time small crystalline needles had formed.
Structure of 1 RS : 1 RS (5 mg) was dissolved in 1 mL EtOH and 3 × 0.5 mL aliquots of water (1.5 mL total) was added to the vial.The resulting translucent solution was then heated until transparent and allowed to cool to room temperature.The vial was left undisturbed for three days after which time plate shaped crystals had formed.

Materials preparation.
HH-gel and TF-gel: typically, 0.5 mL of 1 %wt.solution of 1 RR or 1 SS in 1,4-dioxane was prepared in a glass vial heating a gelator suspension for 5 minutes at 95 ºC in a heating bath.The solution was then thermostatted at 25 ºC or 70 ºC (for HH-gel or TF-gel respectively) soaking the vial in the turned off ultrasound bath for 5 minutes.Then, ultrasound (ultrasonic cleaner bath VWR USC 300 THD/HF) was applied for 60 seconds at 25 ºC or 70 ºC (for HH-gel or TF-gel respectively), and it was left still to gel at room temperature for at least 12 hours.
LC droplets: the clear solution of 1 RR or 1 SS prepared as above was left still at room temperature for one week.The LC droplets started to be visible to the naked eye after two days.When the LC droplets were required for further experiment, they were taken by detaching them from the walls of the vials with a cat whisker and cautiously removing the mother liquors.
LC-gel: the LC droplets suspension prepared above was sonicated at 25 ºC for 10 seconds and left still to gel for at least 1 hour.

Circular Dichroism (CD) and absorbance.
The CD and absorbance spectra were simultaneously recorded using a Jasco J-810 spectropolarimeter.Spectra were obtained at 20 ºC from 215 to 300 nm with a 1 nm step, 8 seconds of digital integration time, 5 nm of bandwidth and a scan speed of 50 nm•min -1 .All spectra obtained were treated subtracting the spectra of pure 1,4-dioxane measured using the same cuvette and conditions, and the.The gelation solutions were prepared in glass vials using the protocols described above and were injected inside 0.1 mm path length quartz cells (Hellma 0.1 mm quartz Suprasils(R)) immediately after sonication.For kinetic experiments, the gelation solutions were introduced in the cuvette already placed in the spectropolarimeter and were left to thermostatize for two minutes before recording spectra every three minutes for LC-gel gelation and every five minutes for the gelation of HH-gel and TF-gel.Except for kinetic experiments, all results are expressed as the average of three accumulations.The baseline of CD spectra was normalized by subtracting the ellipticity measured at 300 nm of the first scan where no absorbance was detected.The HT measured for all the experiments was lower than 400 V at wavelengths higher than 220 nm.

SEM images.
To prepare SEM samples of xerogel dried to the atmosphere, small gel portions were taken with a spatula and put on silicon wafers, left to dry for 1 day and coated with 3 nm of platinum using a Cressington 328 Ultra High-Resolution EM Coating System.The SEM samples of freeze-dried xerogels were prepared freeze-drying the gel, taking a portion with a spatula and put on an adhesive surface, and coated with 10 -13 nm of gold using a Cressington 108 Auto sputter coater.The images were obtained using a Carl Zeiss Sigma 300 VP FEG SEM microscope, operated at 5 kV using an inlens detector.

Rheology.
Rheometry was conducted on a TA Instruments AR2000 rheometer equipped with a peltier plate, using a 25mm rough top plate geometry.To avoid damage to the formed gel, samples were placed in a well plate on the rheometer immediately after sonication and left to gel overnight at 20 °C, covered with a sealed watch glass to avoid solvent evaporation.Measurements were then conducted at a temperature of 10 °C to limit solvent evaporation.Stress sweeps were conducted to determine the limit of the linear viscoelastic region (LVR) and yield stress, at a frequency of 10 rad s -1 .Frequency sweeps were conducted at an appropriate stress within the LVR.The yield stress was given by the end of the linear viscoelastic region, calculated as the point of continuous reduction in G′ of over 5% from a 5-point moving average of the preceding points.
A generalized Maxwell model 3 was fitted to the frequency sweep data 4 in RepTate: 5 (1) (2) Where n is the number of Maxwell modes used to fit the data, equally distributed across a logarithmic scale between and .G i is the modulus of the mode, and τ i the characteristic relaxation time of the mode (given by the inverse of the frequency).The number of modes used was determined by minimising the product of the sum of the squared errors, χ 2 , and the number of fitting parameters: χ 2 (2+n).The elastic shear modulus, G, is determined by summing G i .

Small-Angle Neutron Scattering (SANS).
Each gel sample was prepared as described above, except immediately after the sonication step the sample was pipetted into a titanium demountable cell outfitted with quartz windows and then loaded into the instrument to begin data collection.The elapsed time between the end of the sonication step and the beginning of each scan for each gel is shown in Table S5 in Supplementary Section 6 below.SANS was conducted on the GP-SANS instrument in the High Flux Isotope Reactor at Oak Ridge National Laboratory (Oak Ridge, TN, USA).Three instrument configurations were used to obtain a full q-range of 0.0015 Å -1 < q < 0.81 Å -1 , where q = (4π/λ)sin(θ/2) and is the scattering vector, λ is the neutron wavelength, and θ is the scattering angle.1) a sample-to-detector distance (SDD) of 19 m and λ=12 Å for the low-q data, 2) a SDD of 7 m and λ=4.75 Å for intermediate-q data, and 3) a SDD of 1 m and λ=4.75 Å for the high-q data.The wavelength spread Δλ/λ = 13% for all configurations.To obtain good scattering statistics, the low-q data was collected for 30 min, mid-q data was collected for 20 min, and high-q data collection was 10 min.Jupyter Notebook developed by ORNL was used to reduce the data.All measurements were conducted at room temperature (22°C), and the scattering intensities were calibrated by using porous silica as a secondary standard.The data were corrected for empty cell scattering, sample transmission, sample thickness, and detector sensitivity. Supplementary Section 3: Gelation screening Procedure: In small glass vials, 5 mg of solid gelator were weighted and 0.5 mL of the desired solvent were added to reach a 1% (w/v) final concentration.In case the gelator was soluble, the solution was left still for one week observing its evolution.In case the gelator was insoluble at room temperature, heat was applied until the solvent boiled.If the solid could not be dissolved by heating, that solvent was discarded for further experiments, but in case it was soluble, the solution evolution was observed for a week.All solvents where gels appeared after obtaining the solution were discarder for further experiments.If no gels were obtained after one week, the experiments were repeated sonicating the freshly prepared solution.In case gels were obtained (sonogel), no further experiments were performed.In case a precipitate appeared in a solvent after just dissolving, or dissolving and sonicating, that solvent was discarded.In all cases where no gel or precipitate was observed, the experiments were repeated increasing the gelator concentration to 2% (w/v) and if after that no changes were observed, it was considered that the gelator is soluble in that solvent.All results obtained are summarized in Table S2.

EASY-ROESY
While the variable temperature suggest that the minor and major conformers interconvert, this is not a strict proof of conformer interconversion.An EXSY run in the form of an NOESY or a ROESY can indeed prove that the minor and major sets of signals interconvert, and it can identify the pairs or interconverting signals.Conformational and proximity effects cannot be separated using a NOESY (see below) when the correlation time is smaller than the Larmor frequency, as it was the case since the nOe cross-peaks showed the same phase as the diagonal (Main text Figure 4); however, these effects can be separated using a ROESY (an EASY-ROESY in this case).In a ROESY, cross-peaks due to exchange (chemical or conformational) have the same phase as diagonal peaks; cross-peaks due to TOCSY effects (due to scalar couplings) have the same phase as the diagonal.Here contributions due to scalar couplings were identified using a COSY as a reference (this cannot be done using TOCSY as a reference).The ROESY spectra (Figure S5) was employed to discern which cross-peaks in NOESY were due to proximity.
EASY-ROESY experimental conditions: four transients, 128 steady state transients, and 2,048 complex data points covering 6 kHz were used in both dimensions.Five hundred and twelve complex points were acquired to sample the indirect dimension.The mixing time was 200 ms.The tilt angle was 50.01.The repetition time was 1.68 s, of which 0.68 s comprised the acquisition time.

Estimating inter-proton distances using NMR
Once the exchange contribution was identified and discarded, we used proximity peaks (nOe) in the NOESY spectrum to roughly estimate inter-proton distances (Main text Figure 4).The integral of a nOe cross-peak is proportional to 1/d -6 where d is the distance between two protons.The nOe cross-peak integral between two peak whose distance is known can be used to estimate the proportionality constant.
We used the cross peak between H b -H c for this purpose (the distance is estimated to be 2.340 Å).
In passing we note that this nOe cross-peak coincides with an anti-phase COSY peak, as is usual in these cases; however, the integral of the COSY cross peak should be close to zero, as the signal has an anti-phase character.Nonetheless, distance estimations obtained using NOESY peaks should be used in addition to other evidence, as NMR detects contributions (or the lack of) from the ensemble of conformations in solution, especially when molecules are not rigid.
NOESY Experimental conditions: Four transients, 32 steady state transients, and 2,610 complex data points covering 6.1 kHz were used in both dimensions.Five hundred and twelve complex points were acquired to sample the indirect dimension.The mixing time was 300 ms.The repetition time was 3.4 s, of which 0.4 s comprised the acquisition time.COSY cross-peaks were attenuated using the Keeler-Pell zero-quantum filter.
The main nOe cross-peaks are detailed in Table S4.Every cross-peak has been compared with the two different conformations observed from the single crystal X-ray diffraction (SC-XRD) structures (Figure S4).In conformation A, both halves of the molecule are equivalent, but this does not happen in the bent conformation B. To make the analysis simpler, an arbitrary representation of the molecule has been chosen to assign labels to the hydrogen atoms.In this representation, two halves with different labels can be distinguished: 'left' and 'right' sides (see Figure S4).Table S4.List of nOe cross-peaks with their estimated distances (see text) between protons in Å.A colour code has been used to indicate whether the estimated distance could have been produced by a particular conformer: light green -likely; turquoise -'not impossible'; red -'unlikely'; light blue -uninformative or trivial cross-peaks.
*The distances only include distances between protons in the more abundant conformer. There are several cross-peaks considered 'non-informative' because they are likely to be observed in all conformation, for example correlations: H a - H e , H d -H e , H d -H f , H e -H f , H f -H g , H g -H h and H h -H i .These nOe cross-peaks have been discarded for further analysis.
 There are several unexpected weak nOe cross-peaks (H a -H i , H d -H h , H d -H i , H e -H g , H e -H h and H e -H i ); they are unexpected because they would have to be generated by hydrogens spatially far apart (if they were in the same molecule).It is likely that these nOe are produced by interactions between different molecules.If the linear conformation A arrangement observed in the crystal is considered, the shorter distances between these protons would be 2.3 Å (H e -H g ), 5.7 Å (H d -H h ), 4.1 Å (H e -H h ) and 'too far' (H d -H i , H e -H i and H a -H i ).Except for H e -H g , all these nOe cannot be observed in that arrangement, so, it can be concluded that the molecules are assembled in solution and that this is not like the conformer A.
 If the H f -H h cross-peak is generated by an intramolecular interaction, it favours the B conformation because it would be observed in the 'left side' but unlikely to be detected in any linear conformation.The high intensity of the H f -H h cross-peak may imply that the major conformation in solution is a bent one in a shape similar to the 'left side' of conformer B (Figure S5c). H f -H i cross-peak observed is weak which means that maybe it is of intermolecular nature, but if the bond configuration of 'left side' of the B conformer is considered, this interaction is likely to be observed in certain bent conformations.If an intermolecular interaction in the A conformation is considered, the shortest distance would be 5.3 Å, which is too long to be observed.As it is impossible to be detected as an intermolecular or intramolecular interaction in A conformation, it is concluded that the presence of H f -H i cross-peak discard A conformation as the major conformation in solution.2.9

Liquid crystal formation as seen by solution-state NMR
Non-sonicated samples of 1 RR dissolved in 1,4-dioxane (deuterated or not) develop liquid-crystal droplets, typically after five days.During this time, the frequency of hydrogen atoms bonded to nitrogen shifts to lower frequencies, while hydrogen atoms attached to carbon shift to higher frequencies except for the methylene hydrogens (Figure S5).These shifts have been descried in self-assembled systems in solution when concentration is increased as resulting from hydrogen atoms bonded to carbon being affected by ring-currents, so that they would be unshielded; then, their signals shift to higher frequencies. 9Concomitantly, a shielding of hydrogens bonded to nitrogen atoms would be observed. 10The opposite phenomenon is observed in the 1,4-dioxane -1 RR system; it seems that gelator molecules disappear from solution as they form liquid crystal droplets.

Diffusion Ordered SpectroscopY (DOSY).
We used a double echo-stimulated, convection compensated pulse sequence, using bipolar pulsed gradients (the Varian' Dbppste_cc implementation) to estimate the diffusion coefficient of the molecules in solution.The non-uniformity of the field gradients was calibrated, and the results used to process the data.
Twenty pulse amplitudes ranging from 1.9 to 27.6 G cm -1 in equal steps of gradient squared were used.Sixteen transients, four steady state transients, and 32768 complex data points covering 9.6 kHz were used.The diffusion-encoding pulsed gradient duration was 2 ms.The diffusion time was 200 ms.The gradient stabilisation delay was 2 ms.The repetition time was 5.4 s, of which 3.4 s comprised the acquisition time.The results were analysed using mono-exponential fittings, although the first two points were discarded due to signal abnormalities.The unbalancing factor (alpha) was 0.15.The analysis of the DOSY can be found on the main text.The sonication of the LC droplets was also tested at high temperature.A gel appeared almost instantly upon starting sonication to give highly heterogeneous gels.The resulting material was impossible to load into a syringe and transfer into the CD cuvette for further experiments.The morphology of the microstructure of the heterogeneous gel obtained was observed under SEM and same kind of fibrils similar to the LC-gel obtained sonicating at 25ºC, were observed, Figure S18.

Supplementary Section 6. SANS study
The SANS data was fitted to 1-level and 2-level Beaucage model.The generalized summed model used to fit the data is as follows: 3 ) +   exp ( - 2  ,( + 1) 2 3 )( 1  S5: Timeline of when each low-, mid-and high-q SANS measurement for each time point of each gel began after the gel samples were finished sonicating, in terms of h:min.A time point was considered complete after its high-q measurement was complete: i.e.HH-gel time point 1 was completed when the gel was 1 h 17 min old.

Figure S4 .
Figure S4.Variable temperature NMR.(a) 1 H NMR spectra acquired at temperatures ranging from 25 to 85 ºC.Increasing the temperature increases the frequency of the interconversion between conformers such that only an average conformation is seen at 85 ºC.The process is reversible, as proven by comparing the spectrum acquired at 25 ºC before heating up the sample and after cooling it down to the same temperature (b).

Figure S5 .
Figure S5.600 MHz EASY-ROESY 8 of 1 RR in 1,4-dioxane-d 8 .Cross-peaks due to spatial proximity (rOe, red peaks) have the opposite sign than the diagonal (blue peaks).Cross-peaks resulting from exchange or TOCSY have the same sign than the diagonal (blue peaks).

Figure S6 .
Figure S6.(a) Labelled representation of the linear conformation A of 1 RR obtained from the diffraction of needle-shape ethanol solvate crystals.(b) labelled representation of the bent conformation B obtained from the diffraction of plate-shape acetonitrile solvate racemic crystals.In the representation, an arbitrary distinction between both sides of the methylene group ('left' and 'right' sides) has been done using different labels for the hydrogen atoms.(c) Schematic representation of the different possible configurations of a

Figure S12 .
Figure S12.Handedness of hyper-helical fibres of HH-gel in function of gelator isomer.

Figure S13 .
Figure S13.Zoomed SEM image of the TF-gel xerogel obtained from dioxane solution.

Figure S14 .
Figure S14.XRPD patterns of (top to bottom) air dried xerogels of 1 RR from nitrobenzene (yellow), acetonitrile (purple), propan-1-ol (green), chlorobenzene (blue) and 1,4-dioxane (red) and the calculated XRPD pattern from the 1 RR ethanol solvate (black).The 1,4-dioxane xerogel closely matches the ethanol solvate pattern, however this appears to be as a result of solid state conversion during the drying process.The acetonitrile, chlorobenzene and propan-1-ol xerogels match the xerogel phase obtained from freeze

Figure S15 .
Figure S15.Additional CD spectra.(a) CD spectra HH-gels of 1 SS and 1 RR .(b), CD of HH-gel and TF-gel at the initial stage of the gelation just after applying ultrasound in comparison with the 1 SS 1,4-dioxane solution.

Figure S18 .
Figure S18.SEM images of the heterogeneous, difficult to handle gel from attempts to form LC-gel at elevated temperature.

- 3 o 2 
The first term is the power law model.The TF-gel analysis did not require this term. The second term is the unified exponential/R g model for N structural levels.For the LC-and TF-gels, N = 2.o For HH-gel up to 11 h 13 min, N = 2. o For HH-gel from 14 h 57 min to the end, N = 1. The third term is the incoherent background, accounting for noise. Parameters: o The scattering vector  = ( 4  )sin  Neutron wavelength λ  Scattering angle θ o A, G i , and B i are scaling factors.o R g is the radius of gyration.o (-)Power and Pow i are the exponential intensity decay Table

Table S2 .
Full gelation screening of 1 RR and 1 SS at 1% (w/v) in a range of solvents.IS = insoluble, X = crystallisation, P = precipitation, S = soluble, PS = partially soluble, SH = soluble only when heated, HG = heterogeneous gel, G O = opaque gel, G T = translucent gel, G C = transparent/clear gel.An asterisk indicates where sonication was required for gelation to occur.The symbol † indicates where a result obtained at 2% (w/v).

Table S6 :
Fitting parameters of Power Law + 2 Level Beaucage model fitted to each time point for each gel.The gel age is the time after sonication at which the data was collected.The HHgel only required a 1 Level Beaucage model for the last 4 time points, so the second level's set of parameters were not used.Likewise, the TF-gel did not require a Power Law term to be fitted, so those parameters were not used.The numbers in parentheses represent the uncertainty in the last significant figure of each value.