Anisotropic polymer nanoparticles with controlled dimensions from the morphological transformation of isotropic seeds

Understanding and controlling self-assembly processes at multiple length scales is vital if we are to design and create advanced materials. In particular, our ability to organise matter on the nanoscale has advanced considerably, but still lags far behind our skill in manipulating individual molecules. New tools allowing controlled nanoscale assembly are sorely needed, as well as the physical understanding of how they work. Here, we report such a method for the production of highly anisotropic nanoparticles with controlled dimensions based on a morphological transformation process (MORPH, for short) driven by the formation of supramolecular bonds. We present a minimal physical model for MORPH that suggests a general mechanism which is potentially applicable to a large number of polymer/nanoparticle systems. We envision MORPH becoming a valuable tool for controlling nanoscale self-assembly, and for the production of functional nanostructures for diverse applications.

into plates with a size of 10 mm × 10 mm for AFM imaging. Dialysis membranes (molecular weight cut-off = 3.5 kDa) were purchased from Spectra/Por. DMF, DMSO and other chemicals were obtained from Fisher Chemicals and used without further purification. Dry solvents were obtained by passing over a column of activated alumina using an Innovative Technologies solvent purification system.

NMR Spectroscopy
1 H NMR spectra were recorded on a Bruker DPX-400 or HD500 spectrometer with DMSO-d6 as the solvent. The chemical shifts of protons were relative to solvent residues (DMSO 2.50 ppm, CDCl3 7.26 ppm).

Size Exclusion Chromatography (SEC)
SEC data were obtained in HPLC grade DMF containing 5 mM NH4BF4 at 50 °C, with a flow rate of 1.0 mL min −1 , on a set of two PLgel 5 µm Mixed-D columns, and a guard column. SEC data were analyzed with Cirrus SEC software calibrated using poly(methyl methacrylate) (PMMA) standards.

Refractive Index (RI) Measurements
Values for the refractive index increment (dn/dc) of polymers listed in Supplementary Table 1 were determined using a PSS DnDc1260 differential refractometer fitted with a 620 nm laser. Note that in some cases DLS was used to qualitatively assess the solutions of anisotropic particles, but not to quantify parameters such as DH -for these cases multi-angle light scattering was used (see below).

Static and Dynamic Light Scattering (LS) Analysis
LS experiments were conducted using an ALV-CGS3 goniometer where A is the Avogadro number and is the refractive index increment (see below).
2. The normalized scattering intensity autocorrelation function, Zimm plots were constructed using the Berry transformation, as recommended by Andersson, 2 using a first order polynomial fit for both variables to perform the double extrapolation and thereby estimate the intensity weighted radius of gyration, 〈 G 〉 Z , mass average molar mass, ̅ W , and a virial coefficient to represent pairwise interactions amongst particles, A2.

S9
The REPES algorithm was used to determine relaxation rates, −1 ( , ), that were consistent with a diffusion process from the amplitude correlation function.

Small-Angle X-Ray Scattering (SAXS) Analysis
Small-angle X-ray scattering (SAXS) measurements were made using a Xenocs Xeuss 2.0 equipped with a micro-focus Cu Kα source collimated with Scatterless slits. The scattering was measured using a Pilatus 300k detector with a pixel size of 0.172 mm × 0.172 mm. The distance between the detector and the sample was calibrated using silver behenate (AgC22H43O2), giving a value of 2.481(5) m.
Samples were mounted in 1 mm borosilicate glass capillaries.

Transmission Electron Microscopy (TEM)
TEM observations were performed on a JEOL 2100 electron microscope at an acceleration voltage of 200 kV. All TEM samples were prepared on graphene-oxide (GO)-coated lacey carbon grids (400 Mesh, Cu, Agar Scientific), to enable high contrast TEM images without any staining. 4 Generally, a drop of sample (10 µL) was pipetted onto a grid and left for several minutes, then blotted away. TEM images were analyzed using the ImageJ software, and over 100 particles were counted for each sample to obtain number-average diameter Dn (for spheres), length Ln and width Wn (for worms). Volumes of worms were calculated according to volume, S10 V = πWn 2 Ln/4 (Supplementary Equation 12)

Atomic Force Microscopy (AFM)
AFM imaging and analysis were performed on an Asylum Research MFP3D-SA atomic force microscope in tapping mode. Samples for AFM analysis were prepared by drop casting 5 µL of solution (0.1 mg mL −1 ) onto a silicon wafer that had been freshly cleaned with water and ethanol, then activated using plasma treatment to generate a hydrophilic surface.

Confocal Microscopy
Confocal microscopy images were taken using a Zeiss LSM 880 confocal fluorescent microscope. The solution of the assembly being studied (5 µL of a 0.1 mg mL −1 in H2O) was dropped onto a plasmacleaned microscope slide and left to dry overnight. Assemblies tagged with BODIPY-FL dye (green) were excited using a 488 nm laser, while assemblies tagged with BODIPY-TR dye (red) were excited using a 633 nm laser. Both channels were used at the same time to detect the presence of green or red assemblies and overlays were produced.

Micro-Differential Scanning Calorimetry (microDSC)
MicroDSC measurements were performed on a Nano Differential Scanning Calorimeter 602000 (TA instruments) at 3 atm pressure and with a heating and cooling rate of 0.2 °C min −1 . A volume of 400 μL of solution was introduced into the sample and reference capillary cells. 18.2 MΩ cm water was used to baseline all measurements. The instrument was controlled by the DSCrun software. The data was analysed with the NanoAnalyze software. S11

Synthesis of 3-(adenine-9-yl)propyl acrylamide (AAm)
The AAm monomer was synthesised according to our previous work. 1 To a suspension of adenine (3.0 g, 24.2 mmol) in dry DMF (100 mL), 60% NaH dispersed in mineral oil (1.0 g, 25.4 mmol NaH) was slowly added in small portions under a nitrogen atmosphere. The mixture was stirred for 1 h until no gas was produced. The viscous mixture was immersed into an ice bath and N- (3-bromopropyl) acrylamide freshly synthesized (5.4 g, 28.2 mmol) was added dropwise. The ice bath was left in place and the yellow viscous mixture was stirred overnight. The resulting suspension was concentrated under high vacuum at 50 °C to give a highly viscous oil, to which CH2Cl2 was added and the contents mixed by gentle swirling. The CH2Cl2 was then poured off and the process repeated several times, followed by concentration under vacuum. The crude residue was further purified by column chromatography S12 using a mixture of CH2Cl2 and CH3OH as eluent and a gradient from 1:0 to 9:1 * to give a white solid,

Synthesis of 3-benzoylthymine
Following the procedures in a previous report, 1 benzoyl chloride (11.24 mL, 96.8 mmol) and thymine
The residue was purified by column chromatography with a gradient of CHCl3/CH3OH from 1:0 to S14 93:7 † to give a viscous liquid. Ethanol (20 mL) was then added and the solution cooled to −20 °C to precipitate a white solid ‡ , TAm (1.0 g, 70%). 1

Synthesis of Poly(4-acryloylmorpholine) (PNAM39) Macro-CTA via RAFT Polymerization
The procedures were similar to our previous work. 1 The typical procedure was as follows. For PNAM39,

S1.1.i Syntheses of Diblock Copolymers
The typical procedure was as follows.

Self-Assembly of PT in Water
The seed nanoparticles NT were assembled as follows. The copolymer was dissolved in DMF (at 8 mg mL −1 ) and stirred for 2 h at 70 °C. Then an excess of 18.2 MΩ·cm water was added via a syringe pump at a rate of 1 mL h −1 . The final volume ratio between water and organic solvent was about 8:1.
The solution was then dialyzed against 18.2 MΩ·cm water, incorporating at least 6 water changes, to afford self-assemblies NT at a concentration of ca. 1 mg mL −1 .

Batch Addition of PA to NT
The typical procedure was as follows. The diblock copolymer PNAM39-b-PAAm20 PA was dispersed in H2O at 5 mg mL −1 . This was then added to separate solutions of the nanoparticle NT (0.5 mg mL −1 ) with stirring at A:T molar ratios of 0.07, 0.20, 0.33, 0.67, 1.0, 1.33. The molar ratios were calculated according to the Mn determined from 1 H NMR spectroscopic analyses and the polymers' mass concentrations. The mixtures were then sealed and allowed to stir at room temperature for 2 h. The solutions were then characterized by DLS and TEM.

Stepwise Addition of PA to NT
The procedure was as follows: a solution of PA (0.33 molar ratio of A relative to T) was added to the nanoparticle NT solution (0.5 mg mL −1 ) to give short "seed" worms. After 2 h stirring, further PA solution (0.07 molar ratio A relative to T) was added. This process was repeated until A:T ratios of 0.33, 0.40, 0.53 and 0.67 were achieved. Each stage was characterised by TEM and SAXS analyses.

Analysis of Seed Nanoparticles NT
DLS analysis of NT is presented in Supplementary Figure 9.

DLS Analysis of the Transformation Products
The very long worms generated by the morphological transformation process were not amenable to analysis by single angle DLS, since this assumes a spherical particle. However, the dumbbells formed S26 by addition of PA to NT at an A:T molar ratio of 0.20 were small and compact enough for effective single angle DLS analysis, which is presented in Figure S8.
Supplementary Figure 10. DLS analysis of dumbbells (0.5 mg mL −1 ) formed from NT after adding PA at an A:T molar ratio of 0.20. Note that due to particle anisotropy these single angle DLS measurements are presented for qualitative comparison purposes only.

CryoTEM Analysis of the Transformation Products
To rule out the possibility that the morphological transformation products were artefacts observed in dry state TEM, we performed cryoTEM on the same samples. Images are presented in Supplementary   Figure 12, which demonstrate that the structures were not drying artefacts.
Supplementary Figure 12. Cyro-TEM images of (a) spherical nanoparticles NT; (b) dumbbell-like micelles formed by adding PA to NT at an A:T molar ratio of 0.20. Scale bars = 100 nm.

LS and SAXS Analyses of the Transformation Products
The morphologies of the transformation products were investigated using LS and SAXS. The aim of these analyses was to verify that the particles observed by TEM were representative of the bulk sample.  Figure 13c) and found to fit well to the experimental data at high q (Supplementary Figure 13d), whilst some discrepancy in the region 2 ≤ u ≤ 5 might be explained by variation in the length and thickness of the particle's central region.
Supplementary Table 3

Morphological Transformation at Low PA Concentrations
We attempted to perform stepwise morphological transformation of the nanoparticles NT by adding PA in small aliquots (0.07 molar equivalents per addition). However, as shown in Supplementary   Figure 14, this resulted only in swelling and partial disassembly of the nanoparticles, rather than shape change.

Stepwise Transformation of Dumbbells
To test the hypothesis that stepwise growth might be possible once anisotropy had been introduced, we first took a sample of the dumbbells (A:T ratio 0.20) and added PA, then analysed the resulting particles by TEM (Supplementary Figure 15). As expected, the dumbbells elongated to form worms of around 300 nm length, suggesting that stepwise growth would be possible.
Supplementary Figure 15. Schematic presentation and TEM images of (b) worm-like micelles with lengths over 300 nm formed by further feeding (a) dumbbell-like micelles with PA at a total A:T molar ratio of 0.33; scale bars = 200 nm.

Stepwise Growth of Long Wormlike Nanoparticles
Having confirmed that stepwise growth was possible, we moved on to the experiments described in

SAXS Analysis of Wormlike Nanoparticles
To analyse the wormlike nanoparticles by SAXS we adopted the approach described by Pedersen,6 which considers the scattering form factor (PWORM) to comprise two components that describe the cross section (PCS) and global particles shape (PSH),

Analysis of Mixtures of NT with PA Me , PT1 and PS
In order to confirm that strong H-bonding between A and T was necessary to drive the morphological transformation process, we performed a series of control experiments using polymers in which Hbonding was either partially blocked or completely removed. We began by synthesising four polymers:

Blocking of H-Bonding in the Nanoparticle
We then moved on to investigate the effect of partially blocking H-bonding in the nanoparticle. This

Physical Model for MORPH
This subsection contains the technical details related to the physical model for MORPH presented in the main text. The focus is on relevant timescales in the experimental nanoparticle system, comparison with other polymer systems that tend to relax to equilibrium spherical micelles, and on formulating a more complex mathematical model that takes into account both nanoparticle shape and volume, and which reduces to the equation for eccentricity presented in the main text.

Relevant Timescales
There are many timescales relevant to the formation and growth of the nanoparticles. In the main text, we emphasize the timescales τI for insertion of a polymer into the nanoparticle core and timescale τR for the rearrangement and relaxation of the nanoparticle core chains. We assume τI is inversely proportional to the concentration of the polymer (i.e., more polymer will lead to faster insertion). We assume that τR will be determined principally by the bulk properties of the core chains (i.e., the bulk modulus and viscosity) and remain more or less independent of the polymer concentration. These timescales are microscopic, because they arise from the properties of individual polymer chains and chain-chain interactions.
A different approach to the dynamics is a phenomenological model with timescales for the growth of surface area and volume. The surface-area growth proceeds via timescale A (and, intuitively, surfacearea growth should depend strongly on the insertion timescale τI). The volume growth proceeds via a different timescale V , which depends on the relaxation time τR.

Comparison with Equilibrium Phenomena
In addition to these timescales involved in the MORPH process, there are several timescales that describe equilibrium processes and which, in general, enter the physical description for an equilibrium micelle system. However, due to the non-equilibrium nature of the nanoparticles considered in this work, especially the glassiness of the nanoparticle core, these equilibrium timescales are too long to effect the morphological change from spheres to elongated nanoparticles that we observe.
In most diblock copolymer systems that form micelles, the dominant relaxation process for approaching equilibrium is single-chain extrusion (see, e.g., Ref. [7]). In this process, a micelle loses a single chain to the solution, and this chain diffuses and can rejoin a different micelle. This process leads to a broad equilibrium distribution of particle sizes. However, in the system we consider, singlechain extrusion is strongly suppressed by a combination of the glassiness of the nanoparticle core and the reversible H-bonding interactions between thymine and adenine (which are also not sufficiently strong for PA to pull the core chains out into the solution). As a result, polymers from the core do not individually leave the nanoparticle. Instead, any reformation into smaller nanoparticles must proceed via the budding and break-up of parts of the larger nanoparticles. This budding process is energetically costly and not observed during the shape-change step of the process. Instead, budding and break-up is only observed when these processes are strongly driven by the need to accommodate more polymers, i.e., at the end of the process in Figure 6 when the nanoparticle shell can no longer expand to accommodate more polymer insertion. Another potential disassembly process is a collective modulational instability of a cylinder, described in Refs. [8,9]. This process is expected to be as slow as the budding of individual smaller nanoparticles, and may be responsible for the cylinder-to-smallsphere transformation shown in the final stages of Figure 6.
Finally, surface tension of the core-solute interface is responsible for the (viscous) relaxation from anisotropic shapes into spherical nanoparticles. In equilibrium, this process is responsible for the S62 generic spherical shapes of many micelle systems. This relaxation process depends on the relaxation rate of the core. We believe that in our system, the suppression of this process through glassy core dynamics is essential for stabilizing a worm-like nanoparticle shape.

Swelling Dynamics
In this subsection, we proceed to derive the quantitative relations between the microscopic timescales (τR, τI) and macroscopic timescales ( A , V ). We demonstrate the importance of having a thin shell for the MORPH mechanism: the shell thickness effectively rescales the relaxation rate. As a starting point, consider the equations for the growth of nanoparticle volume and surface area : is the eccentricity of the spheroidal shape,

~ 3 (Supplementary Equation 26)
R is the nanoparticle radius, and a is the core thickness. We assume that the volume growth occurs primarily inside a shell of thickness much smaller than particle radius R (although the case ~ can be considered for shells of thickness comparable to the nanoparticle size).
The timescale A for area growth depends on the core surface tension: the larger the surface tension, the slower the area growth rate. In the absence of polymer insertion, the nanoparticle will prefer a spherical shape that minimizes the core-solution interfacial area and the area growth rate at fixed volume would be negative, driven entirely by surface tension. More generally, the surface tension of the core favours shapes that are more isotropic and can present a barrier to the development of anisotropy.

MORPH Dynamics
The phenomenological model that we present for MORPH dynamics results from putting the above ingredients together. The equation for the evolution of the area can be rewritten as The nonlinear terms ignored here can have two effects: stabilizing a dumbbell shape, and (if odd in ) establishing a preference for elongated (prolate) over squished, pancake-like (oblate) shapes.
Glassiness of the core guarantees a slow relaxation time, so the transition from τI −1 < τR −1 (ϵ = 0) to τI −1 > τR −1 (ϵ ≠ 0) can be realized. The relation given in R −1 = V −1 /(3 ) (Supplementary Equation   31 shows the importance of having a thin shell (achieved through a short length of the added polymer core block): the ratio of shell thickness to nanoparticle radius rescales the effective relaxation rate that enters the equation for ∂ t 2 : the anisotropic regime is easier to probe with a thinner shell.

S64
One simple generalisation of the model that may better reflect the complexity of the polymer system is that the polymer insertion and relaxation timescales may be coupled due to many-body effects. As a result, the eccentricity equation may take the more general form, Although this form changes the dynamical scaling, this more general form leaves the phenomenology of this mechanism intact, i.e., similar development of anisotropic behaviour may be observed in systems with quite different rates of insertion and core relaxation.
In this subsection, we have considered the instability from a spherically symmetric seed to an anisotropic ellipsoid (or worm). On the other side of the instability, once an anisotropic particle has been formed, other effects may play a significant role in the dynamics of further growth. For example, the polymers might exhibit preferential insertion from solution into the nanoparticle based on local surface curvature. In addition, if the inserted polymers diffuse along the surface, the diffusion dynamics could also depend on local curvature, leading to an uneven distribution of polymers on the surface. However, our evidence leads us to conclude that the driving force behind the formation and growth of anisotropic particles is the need to incorporate more surface area at fixed volume: if preferential insertion based on surface curvature was the key mechanism, an anisotropic seed would be necessary for anisotropic growth, whereas our seeds start out spherically symmetric.