Varying the Core Topology in All‐Glycidol Hyperbranched Polyglycerols: Synthesis and Physical Characterization

Abstract

In the present study, low molecular weight cyclic polyglycidol is used as a macroinitiator for hypergrafting glycidol and producing cyclic graft hyperbranched polyglycerol (cPG‐g‐hbPG) in the molecular weight range of 10³–10⁶ g mol⁻¹. Linear graft hyperbranched polyglycerol (linPG‐g‐hbPG) and hyperbranched polyglycerol (hbPG) are prepared as reference samples. This creates a family of hbPG structures with cyclic, linear, and star cores, allowing to evaluate their properties in solution and in bulk. The morphology study of the high molecular weight structures using atomic force microscopy revealed a spherical shape for cPG‐g‐hbPG and hbPG, and a cylindrical shape for linPG‐g‐hbPG in the nanometric range. Small angle X‐ray scattering confirmed the compact particle‐like structure of this family of hbPG architectures. Interestingly, the glass transition temperature showed a structure dependence, with cPG‐g‐hbPG having the highest values and hbPG having the lowest values for the same molecular weight. This study is a step forward in the generation of water‐soluble polymers with tailored structure and functionality for advanced applications.

Cyclic polyglycidol‐graft‐hyperbranched polyglycerols using a hyper grafting strategy from a low molecular weight cyclic polyglycidol are presented along with their analogous structures containing linear and three‐armed star cores. The study of this family of hyperbranched polyglycerols in solution and bulk revealed the critical influence of core topology on their properties, advancing the field of water‐soluble polymers and enabling the design of tailor‐made materials.

1. Introduction

The physical properties and biological activity of polymers are determined by their architecture. The ease of tailoring the size, chemistry, structure, and overall properties of polymers through chemical synthesis has contributed significantly to their widespread use in various technological fields. The diverse properties and functionalities that can be incorporated into polymeric systems are motivating the medical sector to use structured polymers in drug delivery, tissue engineering, and biological imaging.^{[ 1 ]} Polyglycidol or polyglycerol (PG) is a polyether widely used in biomedical and pharmaceutical applications due to its excellent biocompatibility and high functionality, which facilitates the attachment of a variety of bioactive molecules.^{[ 2 ]} With the objective of tuning the physicochemical activity of PGs, several architectures have been explored, including linear, dendritic, and hyperbranched structures. Linear chains are obtained by polymerizing a protected glycidol monomer followed by deprotection of the hydroxyl units.^{[ 2b,c,e ]} Dendrimers are produced through a multi‐step synthesis that allows for well‐controlled molecular architectures.^{[ 3 ]} Finally, the synthesis of hyperbranched PG (hbPG) relies on the one‐pot polymerization of AB₂‐type glycidol monomer using a partially deprotonated multifunctional hydroxyl initiator via ring opening multi‐branching polymerization (ROMBP).^{[ 2} , ^{4 ]} When glycidol is initiated directly from multihydroxyl macroinitiator cores via the so‐called hypergrafting strategy,^{[ 5 ]} hyperbranched blocks can be grown and form unconventional topologies. Some examples include the generation of hyperbranched‐hyperbranched copolymers using hyperbranched macroinitiators,^{[ 6 ]} linear‐hyperbranched graft copolymers using multi‐hydroxyl linear macroinitiators,^{[ 7 ]} and linear‐hyperbranched block copolymers using multi‐hydroxyl block copolymer macroinitiators.^{[ 8 ]}

In the final architecture of hbPG, the core topology clearly plays an important role. For example, the commonly used trifunctional initiator 1,1,1‐tris(hydroxymethyl)propane (TMP) for the production of hbPGs^{[ 4} , ^{9 ]} leads to the formation of globular structures with a three‐armed star core and a random distribution of hydroxyl groups along the branches. Hypergrafting of glycidol from linear poly(4‐hydroxystyrene) macroinitiators has been described for the generation of cylindrical structures.^{[ 10 ]} More complex structures such as a dumbbell or core–shell morphologies have been obtained by hypergrafting the terminal polyglycidol moieties of a PG‐PEO‐PG copolymer or the polyglycerol shell of a hydroxyl‐functionalized hydrophobic dendritic polyethylene core, respectively.^{[ 6} , ^{11 ]} These and other core–shell structures based on hbPG have shown unimolecular transport of hydrophobic model dyes into living cells,^{[ 6b ,12]} demonstrating their great potential in polymer therapeutics.^{[ 2d ]}

There are two general methods for synthesizing cyclic dendronized polymers, including the graft‐from route and the graft‐to route,^{[ 13 ]} which are also used to synthesize, in a more general spectrum, bottlebrush polymers with linear or cyclic backbone cores.^{[ 14 ]} The graft‐from approach involves the stepwise growth of dendrons outward from the polymer backbone, while the graft‐to method involves the reaction of preformed dendrons, which have a single reactive group at the focal point, with a cyclic backbone that has complementary reactive groups. Laurent and Grayson used these two methods to synthesize cyclic dendronized polymers based on a cyclic poly(4‐hydroxystyrene) backbone core and polyester dendrons up to the third generation.^{[ 13 ]} We can include a third method, the monomer route, which involves polymerization, cyclization, and branching, as exemplified by the preparation of branched cyclic polyglycidol (bcPG) by the reaction of glycidol with B(C₆F₅)₃ via electrophilic zwitterionic ring expansion polymerization (eZREP).^{[ 15 ]} In this case, the resulting polymer composition is homogeneous, consisting entirely of glycidol monomers, but with random branching.^{[ 16 ]} In addition, the reaction lacks control over the cyclic core size, with the occurrence of polydisperse ring growth through ring fusion events during polymerization.^{[ 15} , ^{17 ]}

Grafting polymers from/onto cyclic polymer backbones to create cyclic graft (comb or brush) polymers and cyclic dendronized polymers can induce significant changes in their physical and chemical properties when compared to their linear graft counterparts.^{[ 18 ]} Some examples include a higher cloud point temperature for a thermoresponsive cyclic brush copolymer formed by a cyclic polycarbonate core and poly(N‐acryloylmorpholine) brushes,^{[ 19 ]} higher blood circulation times in mice for a high molecular weight cyclic brush copolymer based on PEG brushes,^{[ 20 ]} and higher in vitro cytotoxicity against HeLa cells for a DOX‐loaded cyclic brush copolymer based on cyclic poly(2‐hydroxyethyl methacrylate)‐graft‐poly(N‐isopropylacrylamide‐st‐N‐hydroxyethylacrylamide).^{[ 21 ]} Other properties influenced by the grafting onto cyclic polymers include self‐assembly behavior,^{[ 22 ]} drug release capacity,^{[ 21} , ^{23 ]} and chain dimensions in solution^{[ 24 ]} and in the solid state.^{[ 24c ]}

Cyclic PG (cPG) has not been used as a macroinitiator for the hypergrafting of glycidol, primarily due to the challenges associated with its synthesis, a common problem in the production of cyclic polymers.^{[ 25 ]} The synthesis of cPG can be achieved by ring closure of a preformed linear PG, which is associated with the difficulties of efficient intramolecular reaction of the two chain ends, small‐scale production, and the generation of non‐cyclic impurities.^{[ 25} , ^{26 ]} Other synthetic route is based on ring expansion polymerization, which has the advantage of producing cyclic polymers on a large mass scale, although it also produces non‐cyclic impurities.^{[ 25} , ^{26 ]} In this sense, eZREP of a hydroxyl‐protected glycidyl derivative can be a convenient way to generate cPG. Our recent study on the polymerization of tert‐butyl glycidyl ether (tBGE) with B(C₆F₅)₃ via eZREP showed the formation of cyclic chains along with some non‐cyclic impurities.^{[ 27 ]} However, these contaminants were effectively removed using GPC fractionation and a previously implemented click‐scavenging protocol.^{[ 28 ]}

In the present communication, we have used the previous synthesis protocol to generate cyclic poly(tBGE) chains (cPtBGE). A subsequent deprotection reaction led to the formation of cPG. Then, a cPG chain composed of only 5 monomer units of glycidol was used as a macroinitiator for the generation of cyclic PG‐graft‐hyperbranched PG (cPG‐g‐hbPG) structures of different molecular weights in the range of 10³–10⁶ g mol⁻¹ by ROMBP. As reference samples, analogous linear PG‐graft‐hyperbranched PG (linPG‐g‐hbPG) and hbPG were produced. The behavior of this series of samples, composed only by glycidol monomers, with different architectures in solution was studied by GPC with multi‐angle light scattering (MALS) detection and small‐angle X‐ray scattering (SAXS), and the morphology by atomic force microscopy (AFM).

2. Results and Discussion

2.1. Synthesis of Hyperbranched PG Structures

The synthesis of cPG‐g‐hbPG, linPG‐g‐hbPG, and hbPG structures was performed according to the reactions shown in Scheme  1 . The polymerization of tBGE with B(C₆F₅)₃ was performed in solvent‐free conditions following our previous study.^{[ 27 ]} The product was fractionated in a preparative GPC (Figure S1, Supporting Information), and a fraction characterized by M_n = 700 g mol⁻¹ (X_n = 5) and Ð = 1.1 was separated for further synthesis steps (Figure 1a). MALDI‐ToF MS data showed a peak distribution corresponding to a single species with M_obs (C_n; Da) = nM_tBGE + M_Na. These species are attributed to the formation of cPtBGE chains (Figure 1b). The cyclic purity of the fraction was confirmed by performing a propargylation reaction,^{[ 27 ]} where no signal shift was observed in the mass spectrum after the reaction, indicating the absence of hydroxyl‐terminated chains. Removal of tert‐butyl units was performed by reaction with trifluoroacetic acid followed by hydrolysis of the trifluoroacetic groups with an HCl solution in methanol (Figures S2 and S3, Supporting Information). MALDI‐ToF MS peak distribution (Figure 1c), and the changes observed in the isotopic distribution after removal of tert‐butyl groups (Figure S6, Supporting Information), confirmed the formation of the targeted cPG structure. According to the M_n of the cPtBGE precursor obtained by MALDI‐ToF MS and GPC, the hydroxyl functionality of the obtained cPG is 5 (M_n = 500 g mol⁻¹, Ð = 1.2) (Figure 1a). An analogous linPtBGE sample was synthesized by AROP using potassium tert‐butoxide as the initiator, and HCl/MeOH to terminate the reaction (M_n = 1000 g mol⁻¹, X_n = 7, and Ð = 1.3). The removal of the tert‐butyl units was performed with a similar protocol as for cPG, resulting in the formation of linear PG (linPG) with M_n = 700 g mol⁻¹ and Ð = 1.3. MALDI‐ToF MS data of linPtBGE and linPG (Figure 1d,e) and the changes observed in the isotopic distribution (Figure S6, Supporting Information) confirmed the formation of targeted structures (see also the ¹H NMR and FTIR data in Figures S3–S4, Supporting Information). The hydroxyl functionality of the obtained linPG is 9, including the terminal hydroxyl units. The GPC curves of cPtBGE and cPG showed the expected shift in retention time to a higher value compared to those of linPtBGE and linPG, respectively, due to their more compact structures (Figure 1a).

Scheme 1.

a) Mechanism of eZREP of tBGE with B(C₆F₅)₃,^{[ 27 ]} followed by the cleavage of tert‐butyl groups to generate cPG. Synthesis of cPG‐g‐hbPG by ROMBP of glycidol initiated from a cPG macroinitiator. b) AROP of tBGE, followed by deprotection reaction and ROMBP of glycidol to generate linPG‐g‐hbPG. c) Synthesis of hbPG.

Figure 1.

a) GPC (DMF + 0.1% LiBr) of cyclic and linear poly(tBGE) precursors (cPtBGE and linPtBGE) and their hydroxylated derivatives, cPG and linPG, respectively, obtained after removal of tert‐butyl units. M_n and Ð were obtained with PEO calibration curve. Apparent M_n (M_n(app)) is reported for cPtBGE and cPG. MALDI‐ToF MS data of b) c(PtBGE) (Na⁺), c) cPG (Li⁺), d) linPtBGE (Na⁺) and e) linPG (Li⁺).

cPG and linPG were used as macroinitiators for glycidol hyper grafting by ROMBP (Scheme 1). As a reference sample, a series of hbPGs were synthesized starting from the TMP initiator (Table S1 and Figure S8, Supporting Information). Hydroxyls of PGs and TMP were partially deprotonated with NaH followed by the slow addition of glycidol.^{[ 29 ]} Considering that cPG and linPG are insoluble in THF but soluble in DMSO and dioxane, the latter were used as solvents for the synthesis of cPG‐g‐hbPG and linPG‐g‐hbPG, while THF, DMSO and dioxane were used for hbPG. As shown in previous work,^{[ 9 ]} the use of different solvents has a tremendous influence on the molecular weight and dispersity of the resulting hbPGs, while their degree of branching and physical properties do not depend on the nature of the solvent. For this reason, and with the aim of generating hyperbranched PG structures in a wide range of molecular weights, ROMBP was carried out in different solvents using different glycidol/initiator ratios ([Gly]₀/[I]₀). Table 1 summarizes the molecular weight characteristics of crude hyperbranched samples obtained by GPC, and their degree of branching (DB) obtained by inverse‐gated ¹³C NMR (Figure S7, Supporting Information). Figure 2a,c exhibits the GPC data of crude samples exhibiting broad dispersity in all cases. Using DMSO as a solvent conducted to weigh average molecular weights (M_w) in the range of 13–15 kg mol⁻¹ independently of the [Gly]₀/[I]₀ used, except for linPG‐g‐hbPG which showed the highest M_w value (25 kg mol⁻¹) together with a high dispersity value (Ð ≈5.0). It has been reported that self‐initiation of glycidol can occur in DMSO resulting in a broad and multimodal molecular weight distribution^{[ 6a ]} Since the signals corresponding to self‐initiation (M = nGly + cation) appear at identical m/z values as those of cPG‐g‐hbPG in the MALDI‐ToF MS spectra, the analysis of topological impurities in this series of samples was performed on hbPG obtained in different solvents and the results extrapolated to the rest of the hyperbranched structures (Figures S9 and S10, Supporting Information). The results confirmed that self‐initiation occurred in DMSO but not in dioxane. The use of dioxane as an emulsifying solvent allowed us to generate very high molecular weight samples up to M_w ≈10³ kg mol⁻¹, in agreement with previous studies,^{[ 30 ]} but without any control of the molecular weight. Dispersity values of crude samples were relatively high, with few exceptions, likely due to non‐uniform polymer growth. The GPC data showed the presence of low molecular weight peaks in the crude samples, which were successfully removed by fractional precipitation (Figure 2b,d). The DB values, which ranged from 0.5 to 0.6, confirmed that there was little variation in the hyperbranched structure among the samples, and were in agreement with reported data for hbPG and linPG‐g‐hbPG.^{[ 6} , ^{7 ]} Only for hbPG synthesized in THF, a DB value < 0.50 was obtained. In this sample (S1), the structures obtained by TMP initiation were observed only in the highest molecular weight range (Figure S10, Supporting Information), while in the lowest molecular weight range, the MALDI‐ToF MS spectrum was dominated by signals indicative of self‐initiation (Figure S9, Supporting Information).^{[ 4 ]} Therefore, we attribute the low DB value found for this sample to the high content of topological contamination obtained during polymerization. For that reason, sample S1 was excluded from further analysis.

Table 1.

Reaction conditions of cPG‐g‐hbPG (cyclic core, C series) and linPG‐g‐hbPG (linear core, L series). Polymers were obtained with 100% monomer conversion.

Entry	Solvent	[Gly]0/[I]0 [mol mol−1]	VSolvent/VGly [mL mL−1]	Mw a) [kDa]	Ð a)	DB
C1	DMSO	390	0.2	14	3.1	0.58
C2	DMSO	790	0.2	15	2.1	0.59
C3	dioxane	80	1.0	123	6.0	0.51
C4	dioxane	390	1.0	601	3.1	0.58
C5	dioxane	790	1.0	607	4.1	0.50
L1	DMSO	390	0.2	13	4.1	0.60
L2	DMSO	790	0.2	25	5.0	0.50
L3	dioxane	80	1.0	138	5.1	0.56
L4	dioxane	390	1.0	330	1.8	0.60
L5	dioxane	790	1.0	770	1.4	0.50

^a Obtained by GPC using RI‐MALS detection in DMF + 1% LiBr (dn/dc = 0.054 mL g⁻¹).^{[ 31 ]}

Figure 2.

GPC data (DMF + 0.1% LiBr) of a,b) cPG‐g‐hbPG and c,d) linPG‐g‐hbPG. a,c) Crude samples. cPG and linPG precursors are included for reference. b) Fractions of C1 were obtained in a preparative GPC (F1–F4) and first fractions of C3–C5 (F1) were obtained by fractional precipitation. d) Fractions of L2 were obtained in a preparative GPC (F1–F4) and first fractions of L3‐L5 (F1) were obtained by fractional precipitation. *Toluene.

For a more comprehensive characterization of the synthesized hyperbranched structures, we prepared a series of samples with a variety of molecular weights of low dispersity by recourse to fractionation. Figure 2b,d shows representative fractionated samples obtained from the crude material. Samples obtained in DMSO were fractionated by preparative GPC and samples obtained in dioxane were first dialyzed and then fractionated by fractional precipitation. The molecular weight characteristics of the fractionated samples are reported in Table S2 (Supporting Information).

Branched polymers have a denser structure than their unbranched linear analogs. This results in a shift to higher elution volumes compared to a linear molecule of the same chemical structure and molecular weight. The ratio of the relative M_w obtained from the calibration curve to the absolute M_w obtained from RI‐MALS (M_w(CC)/M_w(MALS)) has been used as a qualitative method to characterize the compactness of branched structures.^{[ 24} , ^{32 ]} By plotting M_w(CC)/M_w(MALS) versus M_w(MALS) using either linear polyethylene oxide (PEO) or linear polystyrene (PS) as standard, different trends were observed for the investigated systems, cPG‐g‐hbPG, linPG‐g‐hbPG and hbPG (Figure 3a,b). In the high molecular weight range (M_w(MALS) > 4 × 10⁵ g mol⁻¹) a convergence of all the data points was observed, indicating that the GPC column cannot discriminate between such large structures. However, as the molecular weight decreased, the polymer structures became selectively distinguished by their hydrodynamic volume. For intermediate M_w(MALS) values in the range of 2 × 10³–2 × 10⁵ g mol⁻¹, M_w(CC)/M_w(MALS) clearly shows that linPG‐g‐hbPG > cPG‐g‐hbPG, suggesting that the latter are the most compact. hbPG seems to follow an intermediate trend in this molecular weight range, being similarly compact as linPG‐g‐hbPG at 2 × 10⁵ g mol⁻¹ and becoming as compact as cPG‐g‐hbPG at low molecular weights. Other interesting features noted in this analysis were the different M_w(CC)/M_w(MALS) values obtained using PEO or PS calibration curves. Values close to 1 indicate that the investigated polymer structures have a similar hydrodynamic volume as the standard linear polymer of identical molecular weight. PEO standards yielded values less than 1 in all cases. However, PS standards yielded values close to and greater than 1 in low molecular weight samples, even though the chemical structure between PS and PG is much more different than that between PEO and PG. As the molecular weight increased in the hyperbranched structures as a result of an increase in branch size, the use of both calibration curves led to smaller molecular weight values than the real ones, as expected.^{[ 33 ]} M_w(CC)/M_w(MALS) reached values as low as 0.05 and 0.1 using the PEO and PS calibration curves, respectively, indicating that M_w is greatly underestimated using the calibration curve method. However, the different trends observed in this type of plot are quite useful for validating the formation of hyperbranched structures with different shapes.

Figure 3.

Plot of M_w(CC)/M_w(MALS) ratio versus M_w(MALS). M_w(CC) was obtained with a) PS and b) PEO calibration curve data, M_w(PS‐CC) and M_w(PEO‐CC), respectively. GPC data was obtained in DMF + 0.1% LiBr for fractionated or monodisperse samples. Solid lines are a guide to the eyes.

2.2. Morphology by Atomic Force Microscopy

In a subsequent step, fractionated samples of high molecular weight (M_w ≈1000 kDa, samples C5–F1, L5–F1 and S5–F1, Table S2, Supporting Information) of cPG‐g‐hbPG, linPG‐g‐hbPG and hbPG were examined using atomic force microscopy (AFM) (Figure  4 ). Diluted solutions of polymers in DMF (0.01 wt.%) were deposited onto silicon substrates by spin‐coating, prior to their observation under AFM. AFM topographical images clearly revealed the anticipated alterations in the morphology of the polymeric chains associated with the different macroinitiators used for their synthesis. Figure 4d–f shows the morphology of three isolated polymeric chains corresponding to cPG‐g‐hbPG, hbPG, and linPG‐g‐hbPG, respectively. While hbPG showed a nearly perfect spherical shape (Figure 4e), an elongated rod‐shaped polymeric chain was observed for linPG‐g‐hbPG (Figure 4f). In the case of cPG‐g‐hbPG, a spherical shape was observed, confirming the expected morphology attributed to the short chain length of the cyclic contour and the long size of the branches (Figure 4d). By reducing the molecular weight at the expense of the branch sizes, the morphology of cPG‐g‐hbPG is expected to transition from spherical to doughnut‐like. However, this transition may occur at very low molecular weights, where the molecular diameter is expected to be less than 10 nm, posing a challenge for AFM resolution. In terms of dimensions, topographical profiles presented in Figure 4g–j shows polymeric chains with a diameter of ≈25 nm for cPG‐g‐hbPG and hbPG, and rods with major and minor dimensions of 50 and 15 nm, respectively, for linPG‐g‐hbPG. The height size is in the range of 5–6 nm for the spherical morphologies and 2–3 nm for the cylinders likely attributed to a smaller branch size given that the linear backbone has 9 anchoring points in comparison to 5 of the cyclic backbone. Finally, the AFM adhesion images (Figure 4k–m) revealed not only the shape of the polymer chains but also their similar chemical composition. Thus, all polymer architectures exhibited comparable adhesion force values, which were found to be lower than those of the silicon substrates. These differences in the observed adhesion forces can be attributed to the reduced presence of hydroxyl groups on the surface of the polymer chains compared to the native oxide silicon surfaces.

Figure 4.

AFM topographical images corresponding to A) cPG‐g‐hbPG, B) hbPG and C) linPG‐g‐hbPG (samples C5–F1, S5–F1 and L5–F1, Table S2, Supporting Information). D–F) AFM topographical images and G–J) their corresponding topographical profiles of a single polymer chain. K–M) Adhesion force mapping corresponds to a single polymer chain.

2.3. Structure and chain conformation as seen by Small‐angle X‐ray Scattering (SAXS)

To further analyze the structure and chain conformation of the synthesized hyperbranched PG structures, SAXS experiments in aqueous solutions were performed (Figure  5 ). SAXS data were obtained for the cPG‐g‐hbPG, linPG‐g‐hbPG, and hbPG structures previously analyzed by AFM (samples C5–F1, L5–F1 and S5–F1, Table S2, Supporting Information) at various concentrations in the dilute range of 1–10 mg mL⁻¹. The results show similar scattering patterns with overall intensity proportional to concentration, indicating that the structure is unaltered in this concentration range (Figures S11–S13, Supporting Information). Interestingly, the scattering intensity is rather high at low Q followed by a rather steep decay with a sign of oscillation at higher Q. This indicates rather compact particle‐like structures, in contrast to typical random polymer chains (for comparison see dashed line in Figure 5). In order to investigate the structure in more detail we initially employed different form factors. Trial fits show that a centrosymmetric “fuzzy” sphere model worked best provided that both a cluster (large aggregates) scattering visible at low Q was included. As seen in Figure 5, the model described in the Experimental Section provides an excellent description of data. It should be pointed out that one would perhaps expect a cylindrical cross‐section for the polymers with linear precursor (linPG‐g‐hbPG case). However, possibly since the precursor is rather small (X_n = 7), the particles may appear centro‐symmetric in solution. Moreover, the low Q scattering is dominated by large molecular weight clusters and polydispersity, also observed in the GPC traces, which partially mask the scattering envelope and makes it difficult to distinguish the exavct particle shape. The resulting fit parameters are depicted in Table S3 (Supporting Information).

Figure 5.

SAXS scattering profiles of aqueous solutions of representative fractions of cPG‐g‐hbPG, linPG‐g‐hbPG, and hbPG of high molecular weights (M_w ≈1000 kDa, samples C5–F1, L5–F1 and S5–F1, Table S2, Supporting Information). The solid lines correspond to the model fits using Equation (6). The Dashed line corresponds to a comparison to the hypothetical case of unaggregated random (Gaussian) chains using the Debye form factor (Equation 8.) with R_g = 5 nm.

As seen the overall dimension is similar for all the samples investigated of the order of 5–6 nm but with larger clusters extending to 6–12 times that size according to the number of spheres per cluster, N_clu, obtained in the fits. In conclusion, we see that the hyperbranched PG structures formed are rather compact and large reflecting the branching, as expected from the synthesis. These results are in agreement in the AFM images previously shown (Figure 4).

2.4. Glass Transition by Differential Scanning Calorimetry

The glass transition temperature is plotted for fractionated samples as a function of M_n in Figure  6 . The data were fitted using the Kanig–Ueberreiter equation (Equation 1),^{[ 34 ]} where the constant k was determined to be 0.03 g molK⁻¹ for linPG‐g‐hbPG, which accounts for a very small variation of T_g with molecular weight, in agreement with the literature for hbPG samples.^{[ 35 ]} The high molecular weight limiting value, T_g ^∞ , for these structures was found to be 265 K. Then, the data for cPG‐g‐hbPG and hbPG were fitted by fixing the value of k = 0.03 g molK⁻¹. The resulting T_g ^∞ for cPG‐g‐hbPG was 269 K and that for hbPG was 263 K. The latter is ≈8 K higher than that found in the literature,^{[ 35 ]} which could be due to different drying methods and pan types used in the experiments.^{[ 15a ]}

$$T_g={\left[\frac{1}{T_g^{\infty }}+\frac{k}{M_n}\right]}^{-1}$$ (1)

Figure 6.

DSC T_g measured for fractionated samples of cPG‐g‐hbPG, linPG‐g‐hbPG, and hbPG. For comparison purposes, T_g data of bcPG fractions obtained from our previous publication were also included.^{[ 15a ]} The dashed lines are the fit using Equation (1).

The data show that T_g is influenced by the type of structure, with cPG‐g‐hbPG having the highest values and hbPG having the lowest values for the same molecular weight. Considering that the degree of branching for this series of samples is in the range of 0.5 to 0.6, where variations in T_g of up to 2–3 K can be expected due to changes in the number of branches,^{[ 36 ]} the difference in T_g found in our samples can be attributed to a different packing of polymer segments within the molecular structure. A plausible explanation is that the packing density in cPG‐g‐hbPG is higher than in linPG‐g‐hbPG and hbPG, making segmental mobility more difficult. The segments closer to the (macro)initiator have to fit into less space near a cyclic structure than near a linear or tridentate molecule, and considering the small size of our cyclic macroinitiator, cPG, the chain crowding near it should be considerably high. As the radial distance from the (macro)initiator increases, the packing constraints should become more relaxed, although the steric constraints imposed by the cyclic structure should somehow remain, as observed in the high T_g values of cPG‐g‐hbPG. As known, the segmental mobility in hyperbranched polyglycerol structures is strongly influenced by the number of branching points, hydroxyl groups, and hydrogen bonds.^{[ 36 ]} The latter was found to be responsible for the non‐variation of T_g with molecular weight in hbPG, as hydrogen bonding interactions between hydroxyl groups from different molecules could cross‐link and form larger supramolecular structures.^{[ 35 ]}

Interestingly, when comparing the molecular weight dependence of T_g of cPG‐g‐hbPG with bcPG (Figure 6) a very different trend is observed. The former exhibits a very slight increase in T_g (k = 0.03 g molK⁻¹) and T_g ^∞ = 268 K, whereas the latter has a large increase (k = 0.24 g molK⁻¹) and T_g ^∞ = 269 K.^{[ 15a ]} bcPG structures are likely formed by different core cycle sizes (which have not yet been estimated) and different branch sizes.^{[ 15a ]} Therefore, the segmental mobility in these types of structures will be dominated by the balance between the number of branching points, hydroxyl groups, and hydrogen bonds described for hbPG,^{[ 36 ]} and by the ring size. The smaller the ring size the higher the expected constraints. The lack of control on the ring and branch size in bcPG may be responsible for the large T_g variation with the molecular weight. On the contrary, the more synthetically controlled cPG‐g‐hbPG structures are formed by the same cyclic core and increasing branch size with increasing molecular weight. Therefore, all the molecules are subjected to similar steric constraints near the core. This effect is probably observed in the lower dependence of T_g with the molecular weight compared to bcPG.

3. Conclusion

We have presented the first example of the generation of a cyclic polyglycidol‐graft‐hyperbranched polyglycerol using the hyper grafting strategy from a low molecular weight cyclic polyglycidol containing 5 monomeric units. This strategy allowed us to produce architectures with compact spherical shapes in the high molecular weight range, which should transition to doughnut‐like structures in the low molecular weight range. The high degree of compaction of these structures was demonstrated in the SAXS experiments by a rather steep decay with a sign of oscillation at high Q, and in the GPC data by longer retention times (and therefore lower molecular weights using the calibration curve) compared to analogous samples with a linear core of similar absolute molecular weight. Moreover, the T_g of the cyclic polyglycidol‐graft‐hyperbranched polyglycerol structures exhibited the highest values of the investigated polymer family and a very small variation of T_g with molecular weight, in agreement with the literature for hyperbranched polyglycerols obtained from a trifunctional star‐like core. The higher T_g values observed for the structures containing a cyclic core could be attributed to the generation of a high packing density of chains near the small cyclic core, an effect that prevails even at high molecular weights. This study reflects the importance of the core topology on the final properties of hyperbranched polyglycerol structures.

4. Experimental Section

Materials

B(C₆F₅)₃, t‐butyl glycidyl ether (tBGE), sodium hydride (NaH),1,1,1‐tris(hydroxymethyl)propane (TMP) and glycidol were purchased from Sigma Aldrich. Monomer and solvents were dried over CaH₂, degassed, distilled under vacuum, stored in Schlenk‐flask under Ar atmosphere, and used freshly purified. B(C₆F₅)₃ was sublimated under vacuum at 110 °C and stored in a glove box in a nitrogen atmosphere. A glove box and vacuum line were used to transfer all the chemicals in an inert atmosphere.

Synthesis of Poly(tBGE)

The 12.7 mL of tBGE (90 mmol) were introduced into a flame‐dried round‐bottomed flask under an Ar atmosphere and cooled to 0 °C. Then, 58 mg of B(C₆F₅)₃ (110 µmol) were added to the monomer under magnetic stirring. The reaction was gradually warmed up to 25 °C and left for 24 h under stirring. Then, the reaction was quenched by the addition of 150 µL mL of DMF, dissolved in dichloromethane (DCM), and passed through a column filled with basic Al₂O₃. The solvent was removed in a rotary evaporator and the polymer dried in a vacuum oven at 80 °C. The product was obtained as a transparent viscous paste. Monomer conversion (¹H‐NMR, CDCl₃) = 100%.

Fractionation of Poly(tBGE)

Poly(tBGE) was fractionated in a recycling preparative GPC, LaboAce LC‐5060, from Japan Analytic Industry (JAI), equipped with a refractive index detector. Separation was performed at 30 °C in 2,5HR and 3HR columns connected in series using the recycling mode. HPLC grade CHCl₃ was used as a mobile phase at a rate of 10 mL min⁻¹. 400 mg of poly(tBGE) were solubilized in 4 mL of CHCl₃ and injected into the columns. Three fractions were collected during the second cycle (Figure S1, Supporting Information). The solvent was removed in the rotary evaporator and the fractions dried in a vacuum oven at 80 °C.

Synthesis of Cyclic PG (cPG)

The 50 mg of Faction 2 of poly(tBGE) was dissolved in 2 mL of DCM, and an excess of trifluoroacetic acid (TFA) (200 µL) was added under stirring. The reaction was performed overnight at room temperature. The excess of TFA was removed in a rotary evaporator. The product was obtained as a brownish oil. The product was then dissolved in 2 mL of ethanol, and 1 mL of HCl (0.1 m) was added. The reaction was stirred overnight at room temperature. Ethanol was removed by distillation in a rotary evaporator, the polymer dissolved in 20 mL of water, washed twice with dichloromethane, and lyophilized. The completion of the reaction was monitored by ¹H‐NMR through the disappearance of the tert‐butyl signal at 1.18 ppm (Figure S2, Supporting Information) and the disappearance of the C = O stretching band of trifluoroacetate groups at 1788 cm⁻¹ and the appearance of a broad signal at 3300 cm⁻¹ in the FTIR spectrum, which is attributed to O─H stretching of polyglycidol (Figure S3, Supporting Information). Yield = 89 wt.%.

Synthesis of cPG‐g‐hbPG

In a typical reaction, the cyclic PG macroinitiator (7 mg, 0.019 mmol, X_n = 5) was placed in a round‐bottomed flask and dried at 120 °C under vacuum for 24 h. Then, NaH (1 mg, 0.042 mmol) was added to the flask inside a glove box, followed by the addition of 1 mL of dry dioxane (or 0.2 mL of DMSO). The system was stirred for 30 min at 95 °C under an argon atmosphere. The activation of the hydroxyl groups of cyclic PG was visually recognized by a color change to pale yellow. Then, glycidol (1.0 mL, 1.1 g, 14.8 mmol) was added during a period of 24 h using a syringe pump (0.04 mL h⁻¹). At the end of the reaction, the mixture was quenched by adding a 0.1 N solution of HCl in MeOH and stirred for 1 h. Monomer conversion (¹H‐NMR, CD₃OD) = 100%. The mixture was passed through a column filled by basic alumina and poured into 500 mL of chilled diethyl ether. The resulting polymer was decanted and dialyzed using a dialysis bag with a molecular weight cutoff (MWCO) of 14 kDa (or 1 kDa) in water. The product was dried in a rotary evaporator followed by lyophilization.

Synthesis of Linear PG (linPG)

In a flame‐dried round‐bottomed flask, tBGE (0.58 mL, 4.10 mmol) was dissolved in distilled THF. Then potassium tert‐butoxide (530 µL 1 m in THF) was slowly added inside a glovebox and the mixture was stirred for 24 h at 60 °C. The reaction was quenched by adding a few drops of 0.1 N HCl in MeOH. The mixture was solubilized in DCM and filtrated in a column of basic Al₂O₃, and the solvent was removed using a rotary evaporator. The monomer conversion of the crude sample [¹H‐NMR, CDCl₃] = 100%. Deprotection of the hydroxyl groups by removal of the tert‐butyl groups was carried out in a similar manner to that used for cyclic PG. The completion of the reaction was monitored by ¹H‐NMR (Figure S4, Supporting Information) and FTIR (Figure S5, Supporting Information) spectroscopy.

Synthesis of linPG‐g‐hbPG

In a typical reaction, linPG macroinitiator (4 mg, 0.008 mmol, X_n = 7) was placed in a round‐bottomed flask and dried at 120 °C under vacuum for 24 h. Then, NaH (1 mg, 0.042 mmol) was added to the flask inside a glove box, followed by the addition of 1 mL of dry dioxane (or 0.2 mL of DMSO). The solution was stirred for 30 min at 95 °C under an argon atmosphere. Then, glycidol (1.0 mL, 1.1 g, 14.8 mmol) was added at a rate of 0.04 mL h⁻¹ using a syringe pump. At the end of the reaction, the mixture was quenched by adding a 0.1 N solution of HCl in MeOH and stirred for 1 h. Monomer conversion (¹H‐NMR, CD₃OD) = 100%. The mixture was poured into 500 mL of chilled diethyl ether. The resulting polymer was decanted and dialyzed using a dialysis bag with a MWCO of 14 kDa (or 1 kDa) in water. The product was dried in a rotary evaporator followed by lyophilization.

Synthesis of hbPG

In a typical reaction, TMP (18.8 mg, 0.14 mmol) was placed in a flame‐dried round‐bottomed flask inside a glove box. Then, NaH (1 mg, 0.042 mmol) was added to the flask, followed by the addition of 1 mL of dry dioxane (or 0.2 mL of DMSO or 1 mL of THF). The mixture was stirred for 30 min at 95 °C under an argon atmosphere. Then, glycidol (1.0 mL, 1.1 g, 14.8 mmol) was added at a rate of 0.04 mL h⁻¹ using a syringe pump. At the end of the reaction, the mixture was quenched by adding a 0.1 N solution of HCl in MeOH and stirred for 1 h. Monomer conversion (¹H‐NMR, CD₃OD) = 100%. The mixture was poured into 500 mL of chilled diethyl ether. The resulting polymer was decanted and dialyzed using a dialysis bag with a MWCO of 14 kDa (or 1 kDa) in water. The product was dried by lyophilization.

Fractionation of Polymer Samples

Polymer samples of cPG‐g‐hbPG, linPG‐g‐hbPG, and hbPG were fractionated by fractional precipitation in a methanol/diethyl ether mixture as follows. Approximately 50 mg of polymer was dissolved in 25 mL of methanol. Under stirring, diethyl ether was added dropwise until turbidity appeared (4–7 mL of diethyl ether depending on the molecular weight). The fraction was collected by centrifugation. This process was repeated for successive fractions.

Gel Permeation Chromatography

GPC data were acquired using a Nexera instrument from Shimadzu using a refractive index detector (RID‐20A, Shimadzu) and MALS detector (λ = 663.89 nm, miniDawn, Wyatt) at a temperature of 40 °C. Separation was performed at 50 °C by using a CTO 40C column oven and Polargel‐M Guard 50 × 7.5 mm and Polargel‐M 300 × 7.5 mm, 8 µm, GPC columns. HPLC grade DMF containing 0.1% of LiBr with a flow of 1.0 mL min⁻¹ was used as a mobile phase. The absolute molecular weights were determined using a dn/dc value^{[ 31 ]} of 0.054 mL g⁻¹ and Astra 8.1 software from Wyatt Technology. Relative molecular weights were determined by using calibration curves of polystyrene and polyethylene oxide, and Lab Solutions 5.1 software from Shimadzu.

Matrix‐Assisted Laser Desorption Index‐Time of Flight Mass Spectrometry

MALDI‐ToF MS measurements were performed on a Bruker Autoflex Speed system (Bruker, Germany) equipped with a Smartbeam‐II laser (Nd:YAG, 355 nm, 2 kHz). Spectra of poly(tBGE), cPG, and linPG were acquired in reflectron mode and those of hbPG in both, reflectron and lineal mode. The laser power was adjusted during the experiments. Poly(tBGE) samples were dissolved in THF and PG samples in MeOH at a concentration of 10 mg mL⁻¹. 2‐[(2E)‐3‐(4‐tert‐Butylphenyl)‐2‐methylprop‐2‐enylidene]malononitrile (DCTB) was used as a matrix for poly(tBGE) samples coupled with sodium trifluoroacetate (NaTFA) as cation donor. Alpha‐cyano‐4‐hydroxycinnamic acid (CHCA) was used as a matrix for cPG, linPG, and hbPG coupled with lithium trifluoroacetate (LiTFA) as cation donor for cPG and linPG, whereas NaTFA for hbPG (10 mg mL⁻¹ dissolved in MeOH). hbPG samples were mixed with the matrix and salt at a 10:5:1 (matrix/polymer/salt) ratio. Approximately 0.5 µL of the obtained solution was spotted by hand on the ground steel target plate and allowed to dry in air. Spectra were accumulated and processed using FlexControl (v3.4) and FlexAnalysis software (v3.4), respectively. Peaks were detected in SNAP mode with a signal‐to‐noise threshold of 3.00 before being processed with a Savitzky‐Golay smoothing algorithm (0.05 m/z width, one cycle) and “TopHat” baseline subtraction. External calibration was performed in quadratic mode with a mixture of different PEO standards.

Nuclear Magnetic Resonance

Polymer samples were analyzed by ¹H and inverse‐gated ¹³C NMR. The spectra were recorded on a Bruker Avance Neo 500 at 25 °C. DCM‐d₂ was used as a solvent for poly(tBGE) and D₂O for PG samples. The degree of branching (DB) was calculated from Equation (2).^{[ 4 ]}

$$DB=\frac{2D}{2D+L_{1,3}+L_{1,4}}$$ (2)

where, D, L_1,3 and L_1,4 are the relative abundance of dendritic and linear structures.

Fourier Transform Infrared Spectroscopy

FTIR spectra were measured at room temperature over the range 600−4000 cm⁻¹ in a JASCO 6300 spectrometer using an attenuated‐total‐reflectance (ATR) stage. Each spectrum was collected with a resolution of 4 cm⁻¹ and an average of 200 repetitive scans. The spectra were neither baseline‐corrected nor smoothed.

Atomic Force Microscopy

The 0.01 wt.% of polymeric solutions in DMF were spin‐coated onto silicon substrates at 2000 rpm for 30 s prior to AFM characterization using a vacuum‐free Ossila Spin Coater. The samples C5‐F1, L5‐F1 and S5‐F1, listed in Table S2 (Supporting Information), were analyzed. AFM imaging was conducted using a multimode Veeco AFM equipped with a Nanoscope 6 controller (Bruker) and PNP Pyrex‐Nitride probes (nominal tip radius <10 nm) in PeakForce Quantitative Nanomechanics Mapping (QNM) mode. Images were taken at a scan rate of 1 Hz and 512  ×  512 pixels.

Differential Scanning Calorimetry

DSC measurements were performed on ≈3 mg samples using a Q2000 TA Instrument. All samples were measured in aluminum pans without lids after it was confirmed that the type of pan significantly affects the reproducibility of the data.^{[ 15a ]} The sample was first cooled from room temperature to −100 °C and then heated to 150 °C at 10 °C min⁻¹ (first heating run). Then, samples were cooled back to −100 °C at 10 °C min⁻¹ and finally heated to 150 °C at 10 °C min⁻¹ (second heating run). A helium flow rate of 25 mL min⁻¹ was used throughout. Glass transition temperatures (T_g) were determined from the maximum of the first derivative in the second heating run.

Small‐Angle X‐ray Scattering

The SAXS experiments were performed using the automated BM29 bioSAXS beamline at the ESRF, Grenoble, France. For technical details we refer to the reference.^{[ 37 ]} The data were obtained using an energy of 12.5 keV and detector distance 2.87 m covering a Q‐range (Q = 4π sin (θ/2)/λ, λ is the wavelength, θ is the scattering angle) of ≈0.0047 Å⁻¹ < Q < 0.5 Å⁻¹. The data were calibrated to absolute intensity scale using water as a primary standard. The solution density of polyglycidol (d_p = 1.34 g mL⁻¹), required to estimate the scattering length density/contrast, ρ_p, in the SAXS modeling (see below), was previously reported by some of us.^{[ 16 ]}

Theoretical Modeling of SAXS Data

The scattering from the polymer structures were described as an ensemble of spherical objects with a fraction, f_clu , of “clustered” spheres, i.e., small aggregates. The total scattering can be written as:

$$I{\left(Q\right)}_{tot}=I{\left(Q\right)}_{NP}\times \left[\left(1-f_{clu}\right)+f_{clu}\times S_{clu}\left(Q\right)\right]+I{(Q)}_{blob}$$ (3)

Q is the modulus of the scattering vector, $Q=4\pi \sin (\frac{\theta }{2})/\lambda$, where λ = 1.54 Å and θ is the scattering angle. Structure factor was used for randomly connected spheres using:

$$\begin{array}{c}\hfill S_{clu,N}\left(Q\right)=\left[\frac{2}{1+\frac{\sin \left(QD\right)}{QD}}-1-\frac{2[1-{\left(\frac{\sin \left(QD\right)}{QD}\right)}^{N_{clu}}}{N_{clu}{\left(1-\frac{\sin \left(QD\right)}{QD}\right)}^2}\times \frac{\sin \left(QD\right)}{QD}\right]\end{array}$$ (4)

where, D = 2R, is the distance between spheres and N_clu is the average number of spheres per cluster. In order to allow for non‐integer values of N_clu we used a linear combination N_clu and N_clu +1, as described in Ref. [38]

For the form factor we used that of polydisperse spheres with graded interfaces described by the interfacial width, σ_R, that can be written as:

$$\begin{array}{ccc}\hfill {〈I\left(Q\right)〉}_{NP}& =& \phi \times {\left({\rho }_p-{\rho }_p\right)}^2\underset{0}{\overset{\infty }{\int }}\frac{1}{\sqrt{2\pi {\sigma }_{PD}}}exp(-\frac{{\left(R_p-〈R_p〉\right)}^2}{2{{\sigma }_{PD}}^2}\hfill \\ & & P{\left(Q,R_p\right)}_{NP}d{R}_p\times \mathrm{ex}{\mathrm{p}}^{\left(-Q^2{{\sigma }_R}^2\right)}\hfill \end{array}$$ (5)

here φ is the volume fraction of polymer given by φ = c/d₀, where d_p is the solution density of the polymers and c is the concentration in g/mL, σ_PD is the Gaussian width of the distribution in the radius of the sphere, R_p. ρ_p and ρ₀ are the scattering length density for the polymer and solvent, respectively.

The blob scattering $I{(Q)}_{blob}$ term can be described by a polymer‐like contribution at high Q and is parameterized by a term given by Lorentz‐type contribution:

$$I{\left(Q\right)}_{blob}=\frac{I_0}{\left(1+Q^2{\xi }^2\right)}$$ (6)

where ξ is a “mesh” size of the internal polymer network and I₀ is a prefactor in units of cm⁻¹.

The scattering length density is calculated according to:

$$\rho =\frac{{\sum }_i{Z}_i}{\overline{M/d_p}}{r}_0$$ (7)

where we found 1.04⋅10¹⁰ cm⁻².

As a comparison we also used the well‐known Debye form factor describing the scattering from polymer chains with Gaussian (random) chain statistics:

$$P{\left(Q\right)}_{chain}=\frac{2·\exp \left[-{\left(Q{R}_g\right)}^2\right]-1+{\left(Q{R}_g\right)}^2}{{\left(Q{R}_g\right)}^4}$$ (8)

where R_g is the radius of gyration.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Pagnacco C. A., Alvarez‐Fernandez A., Maestro A., González de San Román E., Lund R., Barroso‐Bujans F., Varying the Core Topology in All‐Glycidol Hyperbranched Polyglycerols: Synthesis and Physical Characterization. Macromol. Rapid Commun. 2025, 46, 2400791. 10.1002/marc.202400791

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.