From nano to the macro: tuning hierarchical aggregation of thermoresponsive PEG/PCL-based polyurethanes via molar mass/composition control

Amphiphilic hyperbranched polyurethanes (HPUs) based on PEG and PCL are promising for several biomedical applications. However, the lack of control over the molar mass and composition hinders a deep understanding of the aqueous self-assembly of HPUs. In this paper, the control over the HPU molar mass and composition was provided by dynamic urea bond-mediated polymerization (DUBMP), enabling a careful evaluation of their aqueous self-assembly by 1H NMR, DLS, and Cryo-TEM. HPUs containing a single PCL block per chain self-assemble into nanoaggregates (Rh ≈ 10 nm) in water up to its cloud-point temperature (Tcp) of 34 °C. On the other hand, HPUs with more than one PCL block per chain self-assemble into nanoaggregates and their clusters below Tcp. In this case, the solution behavior can be tuned by the HPU molar mass. Increasing Mw¯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{{\mathrm{M} }_{\mathrm{w}}}$$\end{document} from 4 to 19 kDa, HPUs of similar composition can form colloidally stable cluster suspensions (Mw¯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{{\mathrm{M} }_{\mathrm{w}}}$$\end{document} = 4 kDa) and phase separate into a denser liquid aggregate–cluster phase (Mw¯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{{\mathrm{M} }_{\mathrm{w}}}$$\end{document} = 7 kDa) or into a highly viscous aggregate-network phase (Mw¯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{{\mathrm{M} }_{\mathrm{w}}}$$\end{document} = 19 kDa). This type of control over the hierarchical aggregation of HPUs was reported for the first time and is interesting for biomedical applications. The control of amphiphilic branched PU molar mass and architechture via isocyanate reversible deactivation provided the control over its aqueous phase behavior, tuning it from micelles to micelle-networks The control of amphiphilic branched PU molar mass and architechture via isocyanate reversible deactivation provided the control over its aqueous phase behavior, tuning it from micelles to micelle-networks

Overall, the self-assembly of amphiphilic polyurethanes is described using the amphiphilic block-copolymer selfassembly theory with considerable success [30][31][32]. Nevertheless, the influence of the isocyanate segments, molar mass, and architecture on the aqueous self-assembly of PEG/PCL-based branched polyurethanes is still poorly understood [4,10,[22][23][24]. Despite the considerable amount of studies on the subject of PEG/PCL-based PUs, the majority of them focus on other important properties such as cytotoxicity [1,4,24], gelation [4,18,21,23,25,33], and so on. Among these, branched PUs are particularly interesting owing to their capacity to form physically crosslinked networks driven by both self-assembly and/or macroscopic phase separation [4,18,23]. The presence of hydrophilic and hydrophobic segments distributed along a branched network facilitates the interconnection between hydrophobic or dehydrated domains and, thus, the formation of physical networks [4,18,23].
In general, amphiphilic block, or segmented, copolymers that contain more than one hydrophobic segment covalently connected with hydrophilic segments tend to form micellar clusters in aqueous solutions depending on the copolymer concentration [34][35][36][37]. This happens due to the presence of different hydrophobic segments from the same macromolecule at different micelle cores, promoting micelle interconnection and leading to clusterization [34][35][36][37]. Although this behavior is well understood for amphiphilic multi-block copolymers, to the best of our knowledge, it has not yet been described for amphiphilic multi-block polyurethanes. The lack of control over the structure and molar mass of amphiphilic multi-block PUs can complicate these studies, as it increases the complexity of the PU self-assembly. Herein, DUBMP is a promising tool for overcoming these synthetic challenges [18][19][20][21].
Another major challenge when it comes to biomaterials is to achieve hierarchically constructs, such as the ones found in natural tissues, by viable non-toxic synthetic approaches [38][39][40][41]. In this paper, we report on the use of tuning the hyperbranched polyurethane (HPU) molar mass and composition via DUBMP to control the self-assembly of such polymers and their hierarchical self-assembly behavior. This control was able to produce stable aggregates, colloidally stable aggregate-clusters, soft viscous aggregate-cluster phases, and highly viscous aggregate networks. The presence of those aggregates and their hydrophobic-effect driven self-assembly were confirmed by combining proton nuclear magnetic resonance ( 1 H NMR), dynamic light scattering (DLS), cryogenic electronic microscopy (Cryo-TEM), and visual assay results of HPU-aqueous solutions below T cp .

Characterization of the precursors
The absolute number average molar mass ( M n ) of PEG and PCL-triol was determined by the method described by 20,42]. Briefly, polymer solutions in CDCl 3 (5 mg mL −1 ) were analyzed by 1 H NMR before and after acetylation with trifluoracetic anhydride. The molar mass of the PEG and PCL-triol is shown in Table 1.

Synthesis and purification procedures
The synthesis of the HPU was described elsewhere [18]. Mixtures of PEG, PCL-triol, DIPA, and IPDI, always with equimolar amounts of DIPA, hydroxyl, and isocyanate groups, and different proportions of PEG and PCL-triol, dependent on the composition, were heated at 100 °C in an SCHOTT ® tube under N 2 flow for different times, yielding HPUs of different molar mass. The PEG and PCL-triol content was varied targeting different compositions [18]. The polymers were purified by the dissolution of the reaction medium in ethanol followed by precipitation in a diethyl ether/petroleum ether solution (50% v/v) and drying under vacuum.
The synthesis of the linear PU based on PEG and IPDI via DUBMP was described elsewhere [20]. Briefly, a mixture of PEG and IPDI, with equimolar amounts of hydroxyl and isocyanate groups, was mixed simultaneously at 25 °C with DIPA at different n DIPA /n -NCO = 1 molar ratio. The reaction mixture was then heated to 100 °C in an SCHOTT tube under N 2 flow for 4 h.

Characterization
The 1 H NMR spectra of the HPU in CDCl 3 were obtained using a Bruker Avance III HD 250 MHz spectrometer. Solutions of around 5 mg of polymer in 600 µL of CDCl 3 were used. The spectrometer was operated with the following acquisition parameters: temperature of 25 °C, pulse width of 11.7 T, pulse delay of 1 s, acquisition time of 3.3 s, and 16 scans with an FID resolution of 0.3 Hz. The 1 H NMR analyses of a 0.3 wt% HPU solution in D 2 O were performed at 5, 12, and 25 °C on a Bruker Avance II 400 MHz spectrometer using the following acquisition parameters: pulse width of 11.7 T, pulse delay of 1 s, acquisition time of 2 s, and 16 scans with an FID resolution of 0.3 Hz.
Size exclusion chromatography (SEC) analysis was performed in tetrahydrofuran (THF) using a Viscotek GPC-maxVE 2001 chromatograph equipped with a Viscotek VE 3580 RI detector, three Shodex KF-806 M columns, and one Viscotek TGuard 10 × 4.6 mm guard column and operating at 40 °C. Samples were prepared at 5 mg mL −1 of polymer in anhydrous THF and analyzed at a flow rate of 1.0 mL min −1 . The number average molar mass ( M n ), mass average molar mass ( M w ), and molar mass dispersity (Ð) were determined using a calibration curve of polystyrene standards purchased from Viscotek with molar masses from 1050 to 3,800,000 Da.
Dynamic light scattering (DLS) experiments were performed on a Zetasizer Nano: Malvern 3600 equipment using a He-Ne laser (632.8 nm) light source and a detector angle of 173°. Silica gel was included in the closed sample compartment for 30 min before the analysis (25 °C) and during the whole experiment to avoid water condensation on the surface of the cuvette when working at temperatures lower than 25 °C. DLS raw data were treated by the constrained regularization method for inverting data (CONTIN) using the Zetasizer ® software, which provides the number-, volume-, and scattering intensity-based hydrodynamic radius (R h ) distributions.
To address the phase behavior of the HPU-aqueous solutions at 2 wt%, a visual assay was performed. HPU-aqueous solutions were prepared by direct dissolution of the dry polymers in water at 5 °C and kept isothermally at 5 °C for 96 h. After that, the tube inversion test was performed. Samples were then heated to 25 and 37 °C sequentially and then cooled down again to 5 °C followed by vortex stirring. Photos of the samples were taken for all the steps of the assay, which was carried out entirely on a Julabo F12 thermostatic bath for temperature control. Table 1 Mass (f x ) and molar (x x ) fraction determined by 1 H NMR, number of PCL-triol (n PCL ) blocks per chain, and the number of PEG blocks per each PCL block (n PEG /n PCL ) per chain, both estimated by 1 H NMR, M n , M w , and Ð determined by GPC, and T cp determined by DLS a Determined by 1 H NMR b Determined for 0.0125 wt% aqueous solutions [18] c Determined for 0.3 wt% aqueous solution Samples for Cryo-TEM analyses were prepared by direct dissolution in deionized water at 5 °C. The HPU solutions at a 3 mg mL −1 concentration were deposited onto a Lacey Carbon 300 mesh copper grid (Ted Pella Inc.) in a controlled environment vitrification system (Vitrobot Mark IV, Thermo Fisher Scientific formerly FEI) at 5 °C and 100% humidity. The samples were quickly plunged into a reservoir of liquid ethane at −165 °C. The vitrified samples were then stored in liquid nitrogen until they were transferred to a cryogenic sample holder (Gatan 626) and examined with a JEOL JEM-1400 TEM (120 kV) at about −174 °C. The phase contrast was enhanced by under-focusing. The images were recorded on a Gatan multiscan CCD and processed with Digital Micrograph.

Results and discussion
Dynamic urea bond-mediated polymerization (DUBMP) was used to produce HPUs of different compositions based on low molar mass PEG ( M n = 584 Da), PCL-triol ( M n = 1238 Da), and IPDI. Results showed improved control over the HPU molar mass and Ð relative to conventional synthetic approaches such as polyaddition in solution [18]. The HPUs were amorphous with a single T g around −37 °C and thermoresponsive, undergoing an LCST-type macroscopic phase separation in an aqueous solution upon heating above the solution cloud-point temperature (T cp ). This process led to the formation of a physically crosslinked hydrogel for the HPUs of higher molar mass ( M w = 49 kDa) and hydrophobic (PCL) content [18]. The increase in molar mass or the PCL mass fraction decreases the T cp , and the globules formed upon heating above T cp are colloidally stable at 0.0125 wt% [18]. Besides, some evidence suggests that the amphiphilic HPUs based on PEG and PCL self-assemble below their T cp [18,21]. The attenuation of PCL signals in the 1 H NMR of an analogous HPU functionalized with anthracene [21] and the abundance of particles with a hydrodynamic diameter (D h ) of 16 nm, as verified by DLS, suggest that hydrophobiceffect driven self-assembly is occurring [18,21].
However, no HPU containing a single block of PCL-triol was produced. Besides, despite evidence of HPU self-assembly into stable nano-structures (R h ≈ 8 nm) at temperatures lower than the T cp , this behavior was solely investigated by DLS over a short period and for semi-dilute (0.0125 wt%) aqueous solutions [18]. In this paper, a comprehensive study of HPU-aqueous self-assembly in higher concentrations (0.3-2 wt%) and the effect of composition, and molar mass on such a behavior is presented. To do so, an HPU containing a single PCL-triol block per chain (HPU_15_6k) was produced by DUBMP as described in the experimental section. The compositions of all HPUs used in this study determined by 1 H NMR, as well as their molar mass determined by gel permeation chromatography (GPC) and the average number of PCL-triol and PEG blocks per chain estimated by 1 H NMR and GPC, are presented in Table 1.
The molar mass distribution curve obtained by GPC and the 1 H NMR spectra of the HPU_15_6k, as well as the equations used for the HPU mass and molar fractions and estimative of the number of PEG and PCL-triol blocks per chain, are presented in Section I-Structural characterization of the Supporting Information. The GPC curves, M n , M w , Đ, 1 H NMR spectra, and mass and molar fractions of the HPU_25 series were previously reported [18], as well as the DLS data used to determine the T cp of solutions of 0.0125 wt% [18]. DLS data used to determine the T cp of 0.3 wt% HPU_15_6K aqueous solutions are presented in Figure S3-Supporting Information (Section II-Aqueous phase behavior).
Below the T cp , nano-structures of R h around 8 nm were observed by DLS for aqueous solutions at 0.125 mg mL −1 of all the HPUs containing more than one PCL-triol block per chain, as previously reported [18]. The HPU-aqueous solutions at 0.3 wt% were analyzed by DLS at 5 °C (below T cp ) at different times after preparation to address the stability of the mixtures. The R h distribution curves are presented in Fig. 1a. Immediately after solution preparation (0 h), solely scattering objects of R h ≈ 8 nm were observed for the HPU_15_6k, HPU_25_4k, and HPU_25_7K, in agreement with what was reported for the 0.01 wt% solutions of HPU_25 [18]. However, HPU_25_19k presented three peaks at R h ≈ 20, 200, and 2000 nm at 0 h. For the HPU_15_6k solution, the peak of R h ≈ 8 nm stays nearly unchanged after 260 h.
For the HPU_25_4k and HPU_25_7k, isotherms of 90 h below T cp leads to suppression of the peak at R h < 50 nm and to the appearance of a second peak at larger R h values (Fig. 1b, c). For HPU_25_19k, the peaks shifted to higher R h values and the peak centered at R h = 20 nm disappeared as well (Fig. 1d). After 260 h at below T cp , only the 0.3 wt% solutions of HPU_15_6k and of HPU_25_4k, which contains 1 PCL block per chain and a mixture of 1 and 2 PCL blocks per chain, respectively, give suitable results for the DLS analysis (Fig. 1a, b). The quality of the results, as well as the temporal increase in R h of the aggregates of HPU_25 series is also verified in the correlation functions (Fig. 1).
It is hypothesized that the structures of 4 ≤ R h ≤ 10 nm observed by DLS for the HPU at 0 h after preparation are related to micelles with a PCL-core and a shell composed of the PEG-IPDI polyurethane segments. The same hypothesis was raised in our previous studies [18,21]. However, at low concentrations (≤ 0.025 wt%), isotherms below T cp do not show change on the DLS R h distribution curves of the HPUaqueous solutions [18,21]. Therefore, the temporal evolution to larger R h peaks in the DLS results of the HPU_25 series ( Fig. 1b-d) suggests the formation of micelle clusters, a self-assembly event highly dependent on the copolymer concentration [34,35]. The formation of micelle clusters is related to the presence of multiple hydrophobic blocks per chain, which leads to micelle interconnection due to macromolecules that present hydrophobic segments located at the cores of different micelles [34,35]. For the HPU_15_6k, aggregate-cluster formation does not occur, and aggregates are thermodynamically stable, probably because this specific composition presents only 1 PCL-triol segment per macromolecule (Table 1), which agrees with micelle formation. The self-assembly for HPU-aqueous solutions below their T cp is schematized in Fig. 2.
Micelle-clusters can interconnect until the formation of micelle networks, which can reach larger sizes (> 5 µm) and, thus, cannot be precisely analyzed by DLS [34,35]. This happens with the 0.3 wt% aqueous solutions at 5 °C of HPU_25_7k and HPU_25_19k that could not be analyzed by DLS 260 h after their preparation due to the presence of large micro-aggregates (R h > 5 µm). A similar aggregate-cluster formation derived from multiple hydrophobic blocks per copolymer chain was verified for random amphiphilic copolymers of PEG and hydrophobic octadecyl segments [37]. It was found that octadecyl segments can occupy multiple micelle cores, promoting hierarchical micelle aggregation and the formation of micelle clusters [37]. Therefore, the most probable hypothesis is that the nanoaggregates observed by DLS are micelles, with a core composed of PCL-triol blocks, and a shell composed of hydrophilic PEG-IPDI polyurethane segments. A scheme of such hypothesis is presented in Fig. 2. Nevertheless, X-ray scattering experiments need to be conducted to obtain quantitative information regarding aggregate internal structure and the confirmation of micelle (core-shell) morphology. The stability of the HPU_15_6k aggregate solutions below its T cp and the reversibility of the macroscopic phase separation were also confirmed by comparing different methods of preparation ( Figure S3b-Supporting information). Preparing the solution by directly dissolving the dry polymer with cold water (5 °C) or cooling the HPU_15_6k dispersion from temperatures higher than the T cp , to temperatures below T cp , leads to the same DLS result of aggregates with R h ≈ 8 nm ( Figure S3b-Supporting information). This agrees with a self-assembly mechanism, and with the hypothesis of micellization [36]. In contrast, the phase behavior of the aqueous solutions of HPU that contains more than 1 PCL block per chain is more complex due to aggregate-cluster formation.
A visual assay of the phase behavior evolution of HPUaqueous solutions at 2 wt% was carried out by keeping the solutions below T cp for different times, followed by different conditions (Fig. 3). As HPU_15_6k contains one PCL unit per chain, the aggregate solutions are stable, and a clear solution is observed below T cp (Fig. 3a, I-V), agreeing with the hypothesis of micellization [36]. On the other hand, HPU_25_4k, which contains a mixture of chains containing  1 and 2 PCL blocks per chain, turns from a clear aggregate solution at 0 h after preparation to a turbid aggregate-cluster dispersion at 24 h after preparation (Fig. 3b, I and II). The cluster dispersion of HPU_25_4k is colloidally stable up to 96 h (Fig. 3b, III).
For HPU_25_7k, a cloudy aggregate-cluster dispersion is observed after 24 h (Fig. 3c, II). After 96 h, two phases are observed, one denser opaque phase richer in aggregate-clusters and a slightly turbid phase richer in nanoaggregates (Fig. 3c, III). For the HPU_25_19k, the same event occurs 24 h after preparation (Fig. 3d, II). This result agrees with the hypothesis of clusterization via micelle interconnection. As HPU_25_19k has a higher molar mass and, thus, a higher number of PCL segments per macromolecule compared to HPU_25_7k, micelle interconnections are more abundant leading to larger clusters that rapidly sediment.
Besides, the viscosity of the denser aggregate-cluster phase increases as the molar mass of the HPU_25 series increases, as verified by the tube inversion test (Fig. 3c, d, IV), probably due to the same reason. For instance, in the HPU containing 2 PCL blocks per chain, individual HPU chains can connect at a maximum of two different aggregates. As the number of PCL blocks per chain increases, the HPU individual chains can connect three or more different aggregates. This inevitably leads to a decrease in the mobility of the aggregate-cluster-rich phase, culminating in a highly viscous behavior for the HPU_25_19k (Fig. 3d, IV) Heating the HPU solutions above T cp leads to turbid dispersions ( Fig. 3a-d, V). This result confirms that below T cp , the clear HPU solution phase of the HPU_25_7k and HPU_25_19k aqueous systems (Fig. 3c, d, III) is rich in aggregates. The presence of aggregates and solvated polymer chains in the lower density phase of the HPU_25_7k and HPU_25_19k aqueous systems is mainly a result of the osmotic pressure. Cooling the dispersions below T cp and vortex stirring turn the HPU-aqueous systems to a turbid dispersion state, Fig. 3b-d, VII, except for HPU_15_6k, Fig. 3a, VII, which returns to a clear aggregate solution state even before stirring. Therefore, the aggregate cluster formation is not reversible, the clusters cannot be fully dissociated into aggregates, and clear solutions cannot be obtained again for the HPU_25 series even upon vortex stirring ( Fig. 3b-d, VII). This suggests that the clear solution state observed immediately after preparing the solutions from the dry HPU_25_4k and HPU_25_7k is a result of the gradual dissolution of the dry polymer into aggregates/ aggregate clusters. Once the polymer concentration becomes high enough, cluster formation takes place irreversibly. Besides, the fully reversible nature of HPU_15_6k LCST macrophase separation agrees with the hypothesis that the nanoaggregates observed below T cp are thermodynamically stable micelles formed upon hydrophobic-effect driven selfassembly [43][44][45].
To provide additional evidence of HPU aggregation, and insights over its mechanism, 1 H NMR measurements were performed for the 0.3 wt% HPU solutions in CDCl 3 , a good solvent for both PEG and PCL blocks, and on D 2 O, a selective solvent for the PEG blocks (Fig. 4). Comparing the spectra for both solvents, Fig. 4, it is clear that there is an attenuation in the PCL and IPDI signals at D 2 O for all HPUs. The PCL signal suppression agrees with the hypothesis of hydrophobic-effect-driven micellization [43][44][45], being a result of the decreased mobility promoted by the highly packed dehydrated PCL chains present in the core of the micelles [43,46]. However, the IPDI signals were also attenuated, Fig. 4, which is contradictory to the fact that IPDI groups are more abundant in the PEG-IPDI segments located in a hydrated micelle shell at D 2 O.
To better understand IPDI signal attenuation, 1 H NMR spectra in both D 2 O and CDCl 3 of a PU containing PEG and IPDI ( M w = 11 kDa), synthesized in a previous study [20] and purified by the same procedure as the HPUs, are presented in Fig. 4d. There is no signal attenuation in D 2 O for this PU, confirming that both PEG and IPDI are both hydrated. Therefore, IPDI signal attenuation in the HPU 1 H NMR spectra in D 2 O is a consequence of the hydrophobiceffect-driven self-assembly, led by PCL-triol entropically favored dehydration. Being so, the attenuation of the IPDI signals in the HPU 1 H NMR spectra in D 2 O is probably related to inter-chain hydrogen bonding present in the shell of the aggregates composed of the hydrated PEG-IPDI polyurethane segments.
Cryo-TEM images of the 0.3 wt% aqueous solutions of HPU_15_6k and HPU_25_7k at 5 °C (T < T cp ) are presented in Fig. 5. Spherical objects with an average radius of 8 nm were observed in the HPU_15_6k aqueous solution at 5 °C, in agreement with the DLS results (Figs. 1, 5a). This confirms the presence of stable spherical aggregates for the HPU_15_6k, which contains one PCL block per chain. On the other hand, Cryo-TEM images of HPU_25_7k aqueous solutions reveal both spherical particles of an average radius of 4 nm, characterized by the 'darker', higher contrast particles, and clusters of such particles with diameters up to hundreds of nanometers (Fig. 5b). Therefore, Cryo-TEM (HPU_25_7k at 0.3 wt%), as well as the observation of a denser phase formation, macroscopic phase separation process (T > T cp ) (Fig. 3), DLS, and 1 H NMR results, confirm the coexistence of aggregates and aggregate-clusters in the aqueous solutions for the HPU_25 series. At the same time, this results supports the hypothesis of micellization and micelle-cluster formation, although to confirm this hypothesis, X-ray diffraction experiments are required, as previously mentioned [44,45]. The narrowness of the histograms of the aggregate radius, obtained by Cryo-TEM (Fig. 3) from the diameter measurement of over 140 particles, agrees with the DLS results. The smaller radius of the HPU_25_7k micelles (4 nm), when compared to HPU_15_6k (8 nm), is a result of the higher PCL content of the HPU_25_7k (Table 1). A higher PCL content leads to smaller hydrophilic segments of PEG-IPDI between each PCL-triol segment, comparing both HPU that contains similar M w of 6 and 7 kDa, respectively, for HPU_15_6k and HPU_25_7k. This yields smaller micelle shells for HPU_25_7k and, thus, smaller micelles, considering that the size of the PCL-triol segments is the same on both HPUs. This type of control over hierarchical self-assembly can be useful for biomedical applications, since it allows tunable nano-structures that can favor cell growth, besides potentially allowing for the insertion of localized nanofeatures, such as drugs, cell-signaling agents, and so on, onto the biomaterial [47][48][49].

Conclusion
For the first time, the effect of the composition and molar mass of PEG/PCL-based multi-block branched PU (HPU) in their aqueous self-assembly was investigated thoroughly. While HPUs containing a single block of PCL-triol selfassemble into stable nanoaggregates in aqueous solutions, HPUs containing more than one block of PCL-triol per-chain self-assemble into aggregate-clusters. Both aggregate and aggregate-cluster formation were confirmed by the combination of DLS, 1 H NMR, and Cryo-TEM for 0.3 wt% HPU-aqueous solutions, with experiments carried over a time span of 260 h after sample preparation. At a higher concentration (2 wt%), a visual assay showed that aggregate-clusters evolve into a denser aggregate-cluster phase for HPUs containing two or more PCL-triol blocks per chain. Moreover, it was verified that the viscosity of this aggregate-cluster phase is increased by the HPU molar mass and, thus, by the increase in the number of PCL-triol blocks per chain, by favoring the aggregate interconnection. The results agree with the hypothesis of micellization and micelle-cluster formation for the nanoaggregates and aggregate-clusters, respectively, although X-ray scattering experiments need to further be conducted to confirm this hypothesis. Therefore, the PEG/PCL-triol-based branched PU is a promising material for several applications, such as scaffolds, and tissue engineering, especially considering that the hierarchical structure control is a major challenge in such fields. Besides, DUBMP being a solvent/metal-free procedure with improved control over the composition, molar mass, and architecture, relatively to conventional PU synthetic approaches, allows the precise design of HPUs to suit this wide application range.
Acknowledgements The authors acknowledge FAPESP (Process No. 2015/25406-5) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior, Brasil (CAPES), Finance Code 001 for the financial support. The authors would like to thank LNNano/CNPEM for the access to the electron microscopy facility and technical support in the execution of the proposal TEM-C2-27165.
Funding Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature.

Conflict of interest
The authors declare that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.